the impact of rainwater harvesting and fertilizer micro

The impact of rainwater harvesting and fertilizer micro-dosing on farm and household sustainability in rural Tanzania

Diana Lucía Escobar Jaramillo

Degree project • (30 hp)

Swedish University of Agricultural Sciences, SLU

Department of Economics

Agricultural, Food and Environmental Policy Analysis (AFEPA) - Master´s Programme

Degree project/SLU, Department of Economics, 1357 • ISSN 1401-4084

Uppsala 2021

Diana Lucía Escobar Jaramillo

Supervisor: Pierre Chopin, Swedish University of Agricultural Sciences,

SLU, Department of Crop Production Ecology

Assistant supervisor: Jens Rommel, Swedish University of Agricultural Sciences, SLU,


Assistant supervisor: Paolo Sckokai, Università Cattolica del Sacro Cuore, UNICATT,

Department of Agricultural and Food Economics

Examiner: Rob Hart, Swedish University of Agricultural Sciences, SLU,


Credits: (30 hp)

Level: (Second cycle, A2E)

Course title: Master thesis in Economics, A2E

Course code: EX0905

Programme/education: Agricultural, Food and Environmental Policy Analysis (AFEPA) -

Master´s Programme

Course coordinating dept: Department of Economics

Place of publication: Uppsala

Year of publication: 2021

Title of series: Degree project/SLU, Department of Economics

Part number: 1357

ISSN: 1401-4084

Keywords: difference-in-differences, East Africa, food security, innovations,

propensity score matching, sustainability indicators, technology

adoption, Tanzania

Swedish University of Agricultural Sciences

Faculty of Natural Resources and Agricultural Sciences


The impact of rainwater harvesting and fertilizer micro-dosing on farm and household sustainability in rural Tanzania

Approved students’ theses at SLU are published electronically. As a student, you

have the copyright to your own work and need to approve the electronic publishing.

If you check the box for YES, the full text (pdf file) and metadata will be visible

and searchable online. If you check the box for NO, only the metadata and the

abstract will be visible and searchable online. Nevertheless, when the document is

uploaded it will still be archived as a digital file.

If you are more than one author you all need to agree on a decision. Read about

SLU’s publishing agreement here: https://www.slu.se/en/subweb/library/publish-

and-analyse/register-and-publish/agreement-for-publishing/.

☒ YES, I/we hereby give permission to publish the present thesis in accordance

with the SLU agreement regarding the transfer of the right to publish a work.

☐ NO, I/we do not give permission to publish the present work. The work will still

be archived and its metadata and abstract will be visible and searchable.

Publishing and archiving

https://www.slu.se/en/subweb/library/publish-and-analyse/register-and-publish/agreement-for-publishing/

https://www.slu.se/en/subweb/library/publish-and-analyse/register-and-publish/agreement-for-publishing/

Food insecurity and poverty are of major concern for farmers in Tanzania, and the technologies

rainwater harvesting coupled with fertilizer micro-dosing promise to aid in easing these burdens,

particularly in a water-limited context. In this study, I performed an ex-post assessment of the impact

of these two innovations in the sustainability of households and farms, in two contrasting regions of

Tanzania semi-arid Dodoma and semi-humid Morogoro, to see if they would be relevant to promote

in the country. The method used accounts for households’ and farms’ characteristics, estimates

sustainability indicators, and uses a difference-in-differences propensity score matching (PSM)

estimator. The results indicate contrary to expectations, that the households in the semi-arid region

of Dodoma are not benefiting from the adoption of the innovations, neither in food security nor in

economic sustainability and even exacerbated the frequency of water conflicts by 7%. On the

opposite, in the semi-humid region of Morogoro, these two innovations enhanced households’

environmental sustainability and food security by increasing these indices by 3% and 10%

respectively. Using aggregated indicators to assess the impact of rainwater harvesting and fertilizer

micro-dosing on economic, social and environmental aspects, was relevant to show that these

technologies have limited benefits on the sustainability of farmers’ households in Tanzania, thus

need to be complemented by policies that promote households’ characteristics associated with better

food security and economic results, such as training for higher levels of education, greater land

security, and promoting the cultivation of cash-crops.

Keywords: difference-in-differences, East Africa, food security, innovations, propensity score

matching, sustainability indicators, technology adoption, Tanzania

Abstract

1. Introduction ............................................................................................................. 11

1.1. Problem description, research question and objectives .............................. 11

1.2. Background ................................................................................................. 13

1.3. Literature review .......................................................................................... 14

2. Materials and methods ........................................................................................... 18

2.1. The methods ................................................................................................ 18

2.1.1. Household characteristics with Sustainable Livelihoods Approach .... 18

2.1.2. Calculation of sustainability indicators and composite indices ........... 19

2.1.3. Estimation of the impact on sustainability by Difference-in-differences

Propensity Score Matching ....................................................................................... 23

2.2. The dataset descriptive statistics ................................................................ 26

2.2.1. Household characteristics ................................................................... 26

2.2.2. Sustainability indicators ...................................................................... 30

3. Results ..................................................................................................................... 32

3.1. Characteristics associated with the adoption of the innovations ................. 32

3.2. Few impacts of the innovations in Morogoro and no influence in Dodoma . 34

3.2.1. More savings in Dodoma and no impact on economic sustainability in

both regions 37

3.2.2. More environmental sustainability in Morogoro and more water

conflicts in Dodoma .................................................................................................. 37

3.2.3. Improved food security in Morogoro and no impact in Dodoma ......... 38

4. Discussion............................................................................................................... 39

4.1. Morogoro: more food security linked to improved environmental

sustainability but no economic results .......................................................................... 39

4.2. Dodoma: no food security or economic improvements motivated by uneven

distribution of water, water conflicts, soil degradation, and poverty traps .................... 40

4.3. Benefits from the method: recognizing impacts on sustainability of more

education and land security in Morogoro and lower cash-crop area in Dodoma ......... 42

4.4. Limitations of the study ................................................................................ 44

5. Conclusions ............................................................................................................ 45

Table of contents

8

Table 1: Description of sustainability indicators ................................................... 20

Table 2: Summary of Morogoro and Dodoma regions environmental and

agricultural characteristics (Sources: Mnenwa et al. 2010; Graef et al. 2014;

World Bank 2020) ..................................................................................... 26

Table 3: Summary statistics of main household characteristics of Dodoma ......... 28

Table 4: Summary statistics of main household characteristics of Morogoro ....... 29

Table 5: Summary statistics of composite indices for Dodoma ............................ 30

Table 6: Summary statistics of composite indices for Morogoro .......................... 31

Table 7: Estimates of the logistic regression for Morogoro and Dodoma ............. 33

Table 8: Balanced sample sizes ............................................................................. 34

Table 9: Average Treatment Effect on the Treated (ATT) of Rainwater Harvesting

and Fertilizer Micro-dosing Adoption in Dodoma, 2013–2016 (After

matching algorithm applied) ...................................................................... 35


and Fertilizer Micro-dosing Adoption in Morogoro, 2013–2016 (After


Table 11: Summary statistics of household characteristics of Dodoma ................ 52

Table 12: Summary statistics of household characteristics of Morogoro .............. 55

Table 13: Summary statistics of sustainability indicators for Dodoma ................. 58

Table 14: Summary statistics of sustainability indicators for Morogoro ............... 60

Table 15: Summary of Balance after matching in Dodoma .................................. 62

Table 16: Summary of Balance before matching in Dodoma ............................... 62

Table 17: Summary of Balance after matching in Morogoro ................................ 63

Table 18: Summary of Balance before matching in Morogoro ............................. 63


and Fertilizer Micro-dosing Adoption in Dodoma, 2013–2016 (Before


Table 20:Average Treatment Effect on the Treated (ATT) of Rainwater Harvesting

and Fertilizer Micro-dosing Adoption in Morogoro, 2013–2016 (Before


Table 21: Average Treatment Effect on the Treated (ATT) of rainwater harvesting

and fertilizer micro-dosing adoption in yield of maize of Morogoro, 2016 -

2013 ........................................................................................................... 72

List of tables

9

Figure 1: The Sustainable Livelihoods Framework scheme. Based on the figure

presented by the Department for International Development (DFID) (1999)

................................................................................................................... 19

Figure 2: Propensity Score Matching implementation steps. Representation inspired

by the Propensity Score Matching steps presented in the paper of Caliendo

& Kopeinig (2008) .................................................................................... 23

Figure 3:Absolute Standardized Mean Differences of matched sample from

Dodoma ..................................................................................................... 64

Figure 4:Distributional Balance for “share of cash crops” covariate before and after

matching in Dodoma ................................................................................. 64

Figure 5:Distributional Balance for “percentage of total area used for cropping”

covariate after matching in Dodoma ......................................................... 65

Figure 6:Absolute Standardized Mean Differences of matched sample from

Morogoro ................................................................................................... 65

Figure 7:Distributional Balance for “education of the household head” covariate

before and after matching in Morogoro ..................................................... 66

Figure 8: Distributional Balance for “land security indicator” covariate before and

after matching in Morogoro ....................................................................... 66

Figure 9:Distributional Balance for “total area managed by the household”

covariate after matching in Morogoro ....................................................... 67

List of figures

11

1.1. Problem description, research question and

objectives

Food insecurity and poverty are a major concern in Tanzania, and farmers are

affected by the lack of technologies for agriculture. Tanzania is one of Africa’s

fastest growing economies with 7% annual GDP growth (The World Bank 2020).

Agriculture generates 25% of Tanzania’s GDP (World Food Programme 2021) and

employs 77% of the working age adults (The World Bank 2020). However, the

levels of inequality have increased, and 36% of the people suffered from severe

food insecurity in 2017 (Food and Agriculture Organization 2021). The

productivity of crops is affected by irregular weather, use of poor technologies and

reduction of available labour force (Mmbaga et al. 2002). By 2018, only 10% of

Tanzania’s cropped area was mechanically cultivated and almost all rain fed (Lana

et al. 2018). And although rainfall shortages affect agriculture, in semi-arid areas

the most important problems are inter- and intra- seasonal variability; historically,

floods have caused 38% of the disasters in Tanzania and droughts 33% of them,

often floods and droughts occurred in the same semi-arid area during the same

season (Hatibu et al. 2006). Climate change is expected to hinder agricultural

activities even more, increasing the frequency and severity of floods and droughts,

influencing in the outbreak of pests and disease, putting pressure on crop failure

and poor yields (Harvey et al. 2014).

Agricultural innovations in small farms are proposed to face these challenges,

however they influence not only labour activities but also the household and the

whole socio-economic and ecological context within which they are implemented,

therefore sustainability assessments should accompany the process of

implementation to evaluate these connections. Environmental themes have

generally received more attention in sustainability evaluations (de Olde et al. 2016).

But for developing countries genuine sustainability must address food security and

income generation immediate needs as well as long-run sustainability (Lee 2005).

1. Introduction

12

The innovations analysed in this study are rainwater harvesting and fertilizer micro-

dosing. Rainwater harvesting in situ by tied ridges accumulates the water when

there is light rainfall, and in case of heavy rainfall distributes the water and reduces

the speed of the water flow within rows (Germer et al. 2021). Fertilizer micro-

dosing involves applying small proportions of fertilizer with the seed at the time of

planting or as top dressing 3 to 4 weeks after the plant emerges (ICRISAT 2021).

Optimized soil moisture and fertilizer have proved to increase crop yields, water

use efficiency and nutrients use efficiency (Chilagane et al. 2020) hence they can

potentially reduce poverty and improve food security (Habtemariam et al. 2019).

In this study I will account for households-farms characteristics under the

Sustainable Livelihoods Approach; evaluate economic, social and environmental

sustainability measurements through indicators calculated before and after the

adoption of the innovations rainwater harvesting and fertilizer micro-dosing; and

determine possible differences between adopters and non-adopters of the

innovations by the Difference-in-differences Propensity Score Matching estimator.

The question to be answered is: what is the impact of the adoption of rainwater

harvesting and fertilizer micro-dosing over the economic, social and environmental

aspects of sustainability of households-farms of Dodoma and Morogoro regions of

Tanzania?

The specific objectives of this research are: (I) define relevant human, natural,

financial, social and physical characteristics of the households-farms in the regions

of Morogoro and Dodoma in Tanzania, and their changes, to assess their influence

on innovations, (II) estimate the impact of the innovations rainwater harvesting and

fertilizer micro-dosing on the overall sustainability of households in the Morogoro

and Dodoma regions in Tanzania, (III) estimate the impact of the innovations

rainwater harvesting and fertilizer micro-dosing on economic, social and

environmental indicators of sustainability in the Morogoro and Dodoma regions in

Tanzania, and (IV) estimate the impact of the innovations rainwater harvesting and

fertilizer micro-dosing in the food security indicator for Morogoro and Dodoma

regions in Tanzania.

13

1.2. Background

Food insecurity is a global concern and in link with poverty in Sub-Saharan Africa,

furthermore, deteriorating environmental conditions are reinforcing this problem,

by reducing the quantity and quality of water and soil available for farming and

affecting agricultural diversity. The United Nations Zero Hunger Sustainable

Development Goal aims to end all forms of hunger and malnutrition by 2030

(United Nations Development Programme 2015). But the population living under

extreme poverty condition in Sub-Saharan Africa reached 42.3% by 2019 (World

Bank 2020), and 57% of Sub-Saharan Africa’s and Southern Asia’s population

cannot afford a healthy diet (UNICEF 2020). Climate change increased the

frequency of extreme weather disasters by three times that of 1970 and 1980

(United Nations Framework Convention on Climate Change 2021), and agriculture

has been significantly impacted. Economic losses due to extreme weather disasters

that damaged crop and livestock production added up to $30 billion in Africa, for

the period 2008 – 2018 (United Nations Framework Convention on Climate Change

2021).

Agricultural research in Africa has oriented towards the production of

knowledge and technology to face these difficulties and make agriculture the

“motor of sustainable economic growth” (Sumberg 2005). Early economic theory

already highlighted the importance of technical and institutional change, for

instance List (1841) critics to Adam Smith’s poor recognition on the influence of

the intellectual abilities on nations’ revenues and not only material capital. In the

same trend the World Bank’s conclusion that investing in knowledge accumulation

is decisive rather than physical capital investment, based on the “New Growth

Theory” that states increasing marginal productivity of knowledge as an input in

production (Romer 1986) (The World Bank 1991). Nevertheless, innovation

systems theory recognises the limited benefits that agricultural technology has had

for poor people in Sub-Saharan Africa. This is attributed to the instability and site

specificity of agriculture and to the failure of the market to allocate technological

resources, because poor farmers are in no economic or risk taking position to invest

in research and development unless the government or other institutions intervene

(Clark 2002).

In this study, the effects of the potentially relevant and easy-to-implement

technologies rainwater harvesting and fertilizer micro-dosing adopted in Tanzania

are evaluated, this is done by calculating indicators that provide information on the

impact of innovations on the three pillars of sustainability in the same analysis. The

innovations are now viewed in the frame of a humankind-nature balanced

benefitting relationship, rather than just aiming to economic goals related to

productivity (Andrade et al. 2020). Some recognized shortcomings of current

14

sustainability studies of agriculture that I target are: not addressing specifically the

multi-functionality of agriculture and ignoring some of the three dimensions of

sustainability, ecological, economic and social, and not contemplating the

interaction and trade-off between sustainability indicators (Bonisoli et al. 2018).

Studying the impact of innovations requires to control for other characteristics

of households that may have evolved independently from the adoption of the

innovations. The framework applied to account for households’ characteristics is

the Sustainable Livelihoods Approach that is adequate for developing countries

where there is no clear separation between household and farms characteristics. The

Sustainable Livelihoods approach considers that people are operating under a

vulnerability situation therefore it states that they manage assets considered poverty

reduction factors (DFID 1999), and classifies the capital into five categories:

financial, physical, natural, human and social (Quandt 2018).

This study uses the Difference-in-differences Propensity Score Matching

method to evaluate the influence of self-selection to adopt the innovations, and

analyse if the characteristics of the individuals adopting the innovations are

responsible for their results on sustainability or if they are influenced by the

treatment. The Difference-in-differences Propensity Score Matching approach is

effective in reducing the time varying bias and selection bias that could arise in

before-and-after comparison (Udagawa et al. 2014). There is a limited literature of

the application of this method to analyse agricultural innovations’ impact on

sustainability of farmers in developing countries; while it is known that the adoption

of sustainable agriculture varies depending on farmers’ socio-economic

characteristics, attitude and beliefs (Comer et al. 1999).

1.3. Literature review

A basis for this research is the Sustainable Livelihoods Approach, which is the

framework used for reporting the households’ and farms’ characteristics which will

be controlled for changes during the period of study and for their connection with

the adoption of the innovations. There is some literature using the sustainable

livelihood strategy framework to evaluate the connection between economic and

social choices of households such as productive activities and assets investments,

and their impact on ecological, social and economic results (Rakodi 1999; Barrett

et al. 2001; Pender et al. 2001; Jansen et al. 2006b). Particularly, Jansen et al. (2006)

made a qualitative and quantitative measurement of livelihood strategies based on

land and labour use in Honduras hillside, with the objective of providing

15

information about potential policies for conservation of the natural resources. The

authors used factor and cluster analysis to cluster farm households based on their

land resources and labour use, and through regression models they analysed the

relationship of livelihood strategies, physical, financial and social capital, with

income per capita, land management and soil conservation practices at parcel level.

The results indicate that security in land tenure and education promote sustainable

land-use practices and income, and that investments to facilitate households-farms’

extension of productive assets should focus on farms that have off-farm

employment because of their higher opportunity costs of labour. The study of

Jansen et. al provides a reference on how the sustainable livelihood approach

captures household-farms’ characteristics, endowments and priorities, so then they

can be connected them to the farms’ outcomes, to assess the effectiveness of support

for adoption of valuable practices. Nevertheless, it is recognized that the

Sustainable Livelihoods Approach needs to be adequately coupled to other analysis

to capture the impact of agricultural systems, in our case to evaluate the impact of

technologies on the social, economic and environmental sustainability, given

adopters characteristics.

For impact assessments of rainwater harvesting and fertilizer micro-dosing, most

studies focus on ecological effects or results on yields (Binder et al. 2010), at the

landscape and field level, while the multiple impacts of these technologies on the

sustainability of households is limited. For instance, Saidia et al. (2019) conducted

field experiments to assess the impacts in yields and land utilization efficiency of

tied ridges water harvesting, fertilizer micro-dosing and intercropping, the

experiment was performed in sub-humid regions of Tanzania. The authors

concluded that combining inter-cropping with water harvesting and fertilizer micro-

dosing can increase the income and food security of small-hold farms in sub-humid

Tanzania. On the other hand, Vohland & Barry (2009) made an evaluation of food

security, income generation, and the ecological impact concentrating on the

landscape scale, because they aimed to evaluate water dynamics beyond field level

in African drylands. The approach applied was the recompilation of literature about

rainwater harvesting, and they also performed a nonlinear regression relating grain

harvest against mean precipitation. The authors concluded that the effect of

rainwater harvesting on landscape functions is positive respect to aquifers recharge

and soil water, while the results for the social and economic sustainability will

highly depend on the involvement of farmers and general communities. Vohland &

Barry suggested that a more complete quantitative analysis at the household level

is required, to measure socioeconomic factors. Studying the impacts of rainwater

harvesting and fertilizer micro-dosing at the household level, connecting the

economic, social and environmental sustainability, would result on a more

16

comprehensive evaluation of these technologies for further recommendations about

their use.

Studies to assess the economic, social and environmental impacts of rainwater

harvesting and fertilizer micro-dosing in Tanzania have focused on expected

results, while to the best of my knowledge no ex-post assessment that includes the

three dimensions has been made. Schindler et al. (2016a) combine stakeholders’

and researchers’ knowledge in an ex-ante (before) impact assessment of agricultural

innovations that include rainwater harvesting and fertilizer micro-dosing. After

applying the impact assessment approaches Framework for participatory impact

assessment and ScaIA-Food Security to consider stakeholders and researchers

evaluations, Schindler et al. conclude that farmers consider indirect impacts not

observed by researchers, that are important to improve the assessments. Some

differences include farmers’ consideration of the risk of lack of rain and chemical

fertilizer application, that would increase yield failure in case of drought even

further, instead of generating the yield increases expected by researchers.

Moreover, farmers envision the increased workload to construct the infiltration pits

that would reduce the field sizes that they are able to cultivate due to limited labour

availability. On the other hand, Graef et al. (2017) focus on an expert-based ex-ante

social, economic and environmental impact assessment of a group of innovations

that include also rainwater harvesting and fertilizer micro-dosing. The authors

highlighted that rainwater harvesting and fertilizer micro-dosing are expected to

have a higher impact in the semi-arid region by improving food diversity, social

relations and working conditions. The ex-ante assessments mentioned provide

expected results of functioning and impact of innovations that will be confirmed or

not in this study.

Regarding the method applied in this research, Difference-in-differences

combined with Propensity Score Matching has been used to analyse the impact of

the implementation of agricultural policies and of the adoption of agricultural

technologies, given panel data and non-random treatment assignment, however as

far as I know it has not been coupled with sustainability indexes before. An

approximation to the use of these methods in the present research is the study of

Dillon (2011), who examined the impact of small-scale irrigation adoption over

households’ consumption, assets and the informal insurance practice of food

sharing between irrigators and non-irrigators, the research is done for northern

Mali. The author used Difference-in-differences Propensity Score Matching

considering that access to irrigation is not random but influenced by households’

characteristics such as family size, education of the household head, assets,

landholdings, and ethnicity. The conclusion reached by the author is that small-

scale irrigation projects have the direct impact of benefitting households’

17

consumption by 27 to 30% respect rainfed and water-recession cultivation in

regions affected by risk of droughts, but also the indirect impact of promoting what

is called informal social insurance that consists of the increase of savings and

sharing within the villages. Läpple & Thorne (2019) used an extension to the

method by applying a Generalised Propensity Score to assess the effect of different

levels of innovativeness of Irish dairy farmers on their economic sustainability. The

authors concluded that innovativeness in general generates higher economic gains,

but not in a linear way, and that highly innovative farmers can still improve their

economic results from further innovativeness. These studies show examples of the

use of the method and what we could expect in food security from the application

of water access technologies, and economic gains from innovations adoption, but

also support the need of doing a context-based analysis that considers regional and

farmers’ characteristics when assessing the impact of agricultural technologies, and

coupling the method with composite indicators appears adequate to evaluate effects

in sustainability.

The literature shows that the Sustainable Livelihoods framework and

Difference-in-differences Propensity Score Matching may be complementary

approaches to identify characteristics of farms and households, and assess their

connection to the effects of agricultural innovations on sustainability. To the date,

an ex-post assessment of the impact of rainwater harvesting and fertilizer micro-

dosing in the economic, social and environmental aspects of sustainability of

farmers’ households in Tanzania has not been done.

18

2. Materials and methods

The data used was collected by a survey conducted in Tanzania as part of a large

trans- and inter disciplinary research project1. The dataset was organized in a panel

with observations per household and consisted of 448 observations for the region

of Dodoma and 444 for the region of Morogoro in the year 2013, and 420

observations for Dodoma and 391 for Morogoro in the year 2016.

Rainwater harvesting and fertilizer micro-dosing were assessed in this study due

to the higher adoption rate and importance that farmers attributed to these

technologies in their food production process, respect to other innovations, as

assessed by Schindler et al. (Schindler et al. 2016b).

2.1. The methods

The process followed in this research to assess the impact of the innovations

rainwater harvesting and fertilizer micro-dosing on farms-households sustainability

consisted on three main steps: (1) definition of the household characteristics

according to the Sustainable Livelihoods Approach, (2) calculation of the

sustainability indicators and composite sustainability index, and (3) estimation of

the impact on sustainability by Difference-in-differences Propensity Score

Matching.

2.1.1. Household characteristics with Sustainable Livelihoods

Approach

Data from a survey for the years 2013 and 2016 was used to classify the

information, calculate ratios, and transform into dummy variables, to present the

household characteristics according to the Sustainable Livelihoods framework. The

Sustainable Livelihoods Approach classifies assets considered as poverty reduction

factors, into five categories of capital: financial, physical, natural, human and social

(Quandt 2018).

1 The Trans-SEC project proposes agricultural innovations as a way to use research and knowledge to face the

need for food security of the rural poor population in Tanzania. For more information see: http://www.trans-

sec.org/

19

Figure 1: The Sustainable Livelihoods Framework scheme. Based on the figure presented by the

Department for International Development (DFID) (1999)

2.1.2. Calculation of sustainability indicators and composite

indices

The sustainability indicators were chosen based on literature about researchers’ and

farmers’ opinion of aspects that represent sustainability, they pertain the economic,

social and environmental pillars of sustainability and use the information available

from the surveys. Thus, I adopted an approach similar to the one by Yegbemey et

al. (2014), who used a participatory method to account for farmers’ and agricultural

extension officers’ viewpoints when selecting the indicators to evaluate agricultural

sustainability. Also, the selection of the indicators in each component of the

sustainability pillars was done following the process proposed by ul Haq & Boz

(2018) to account for site specific features. Table 1 presents a list of the indicators

calculated and includes information about the direction which explains if the

indicator is considered to influence positively in the sustainability component

(“Additive”) or negatively (“Subtractive”).

20

Table 1: Description of sustainability indicators

Pillar Component Indicator Units Direction

Envir

on

men

tal

Soil

management

Quantity of applied

fertilizer Kg / ha Additive

Quantity of animal

manure applied Kg / ha Additive

Crop area perceived

as unfertile % area Subtractive

Crop area with

perceived decrease

in fertility

% area Subtractive

Crop area under

legume % area Additive

Crop area with

residues left on the

field

% area Additive

Area where there is

an intent to invest in

soil fertility

% area Additive

Average tree density Number of trees / ha Additive

Area under erosion

control measures % area Additive

Water

management

Presence of

irrigation Yes – No Subtractive

Rainfall water use

efficiency

Kg / ha / mm of

rainfall Additive

Change in household

water consumption Litres/day Subtractive

Water harvesting Yes – No Additive

Water use conflict % of amount Subtractive

Agricultural

diversity

Tree diversity Number of species Additive

Crop diversity

Number of species Additive

Livestock diversity Number of species Additive

Eco

no

mic

Crop

profitability

Crop gross margin USD / ha Additive

Crop expenditures USD / ha Subtractive

Labour productivity USD / person Additive

Post-harvest loss % amount Subtractive

Profitability Net household

income USD Additive

Stability

High income

fluctuation Yes – No Subtractive

Has savings Yes – No Additive

Reduction of

vulnerability

to shocks

Loss of income due

to shock USD Subtractive

High severity of

shock Yes – No Subtractive

Time to recover after

shock Number of months Subtractive

21

So

cial

Food security

Potential Food

Availability index

(PFAI)2

Ratio Additive

Months of

inadequate food

provisioning

Number of months Subtractive

Food Consumption

Score (FCS)3 Score Additive

Coping Strategies

Index (CSI)4 Score Subtractive

Health

Health insurance

binary Yes – No Additive

Healthy household

members

% of people Additive

Wellbeing

Hours worked Hours/active

household member Subtractive

Perceived

deterioration of

household situation

Yes – No Subtractive

High impact of

income fluctuations

on wellbeing

Yes – No Subtractive

Social capital

Information network Number of sources Additive

% of crops receiving

support from farmers

group

% Additive

Land security

Land title ownership % of area Additive

Secure land % of area Additive

Land use conflict % of area Subtractive

Using these sustainability indicators, composite indices were calculated for each

region, Dodoma and Morogoro. Before this calculation continuous variables were

treated for missing values, by completing the information using the function

imputePCA from the missMDA package on R, which predicts the missing values

through a model based on complete observations. The number of values missing

from the dataset of indicators of year 2013 were 95 out of 41,924, and for the dataset

of year 2016 were 12 out of 34,062. Outliers were caped to control for data

recording errors, by replacing the values above the 95th percentile by the 95th

percentile, and values below the 5th percentile by the 5th percentile. Next, the

2 Potential Food Availability index (PFAI): represents a households’ potential food consumption expressed in

energy equivalents respect its energy needs for a year (Frelat et al. 2016). 3 Food Consumption Score (FCS): the possible range is between 0 and 112 points and represents the frequency

weighted diet diversity (World Food Programme 2015). 4 Coping Strategies Index (CSI): represents the frequency and severity of coping behaviours that households

adopt when they do not have access to enough food, a lower score indicates less food insecure (Caldwell &

Maxwell 2008).

22

calculation of the composite indices consisted of the normalization of the

component indicators, assignment of weights, and aggregation of the component

indicators.

The normalization of the component indicators was done to obtain homogenous

units that enables their comparability and performing arithmetic operations with

them. The method used for normalizing was the min-max normalization (see

Gómez-Limón & Sanchez-Fernandez 2010; Haileslassie et al. 2016; Mutyasira et

al. 2018), considering the minimum and maximum of the whole data of both periods

2013 and 2016. Equation (1), for additive indicators, and equation (2), for

substrative indicators, present the formula used for normalizing.

𝑥𝑛𝑜𝑟𝑚 =𝑥 − min(𝑥)

max(𝑥) − min(𝑥) (1)

𝑥𝑛𝑜𝑟𝑚 =max(𝑥) − 𝑥

max(𝑥) − min(𝑥) (2)

Where xnorm is the normalized value of the observation x, and min(x) and

max(x) are the minimum and maximum observations in the whole 2 years’ sample.

The resulting value will lie between [0, 1], being 0 the least sustainable and 1 the most

sustainable. Then, the normalized indicators were assigned weights which summed

up to 1 per component. Binary indicators were given half the weight of continuous

indicators, following Chopin et al. (2019) who explain that binary indicators

provide less information. Lastly, the composite indices per component were

calculated by aggregating the weighted indicators. The formula applied was the

weighted sum:

𝐶𝐼 = ∑ 𝑤𝑖 ∗ 𝑥𝑛𝑜𝑟𝑚𝑖 (3)

𝑛

𝑖=1

Where CI stands for composite index, n is the number of indicators for the

composite index, w is the weight assigned to the component indicator, and xnorm

is the normalized indicator. Following, composite indices per pillar of sustainability

and a composite index of sustainability, composed by indicators for the three pillars,

were calculated by a weighted sum where each component had the same weight

considering that they had equal importance.

23

2.1.3. Estimation of the impact on sustainability by Difference-

in-differences Propensity Score Matching

The method to evaluate the impact of the adoption of the innovations on the

sustainability of households-farms was Differences-in-differences combined with

Propensity Score Matching, which are commonly applied to analyse the impact of

policies and programmes. Differences-in-differences evaluates the impact of a

treatment on the outcome change over a period by comparing the treated individuals

with the matched control non treated ones. Propensity Score Matching controls for

selection bias, i.e. making the two groups comparable. As defined by Caliendo &

Kopeinig (2008), selection bias occurs when we want to assess the difference in the

outcome of the individuals with and without the treatment, but we cannot observe

both outcomes for the same observation at the same time, either the individual was

treated or not. Comparing treated and non-treated units might result in biased

estimates and this happens when there exist characteristics that affect

simultaneously the outcome and the probability to receive the treatment (Chabé-

Ferret 2015). If that happens, the difference in the outcome may be due to the

difference in the characteristics between the individuals and not in the treatment

itself. This issue often arises in non experimental setting. Therefore, using

Differences-in-difference with Propensity Score Matching looks adequate for this

study, given the impact analysis we are aiming to perform and considering that the

adoption of the innovations was decided by each farmer, that may be prompt to self-

select given their personal and labour conditions.

I followed the procedure applied by Arata & Sckokai (2016) for the Difference-

in-differences Propensity Score Matching method. First the Propensity Score

Matching was performed, as portrayed in figure 2.

Figure 2: Propensity Score Matching implementation steps. Representation inspired by the

Propensity Score Matching steps presented in the paper of Caliendo & Kopeinig (2008)

24

To overcome the selection bias, the Propensity Score Matching matches each

treated individual with one or more non-treated individuals that have similar

observed characteristics called covariates, X. According to Rosenbaum & Rubin

(1983), the probability of being treated P(X) is conditioned on a function of X, and

the matching is based on P(X). The covariates were chosen using a logistic

regression and followed Caliendo & Kopeinig (2008) instruction that outcome

variables must be independent on the treatment and conditional on the propensity

score. The average treatment effect on the treated (ATT) is used to evaluate the

effect of the adoption of the innovations and is calculated as the difference in the

mean outcomes of the treated units and the matched control group.

𝐴𝑇𝑇 = {𝐸(𝑌1|𝐷 = 1, 𝑃(𝑋)) − 𝐸(𝑌0|𝐷 = 0, 𝑃(𝑋))} (4)

Where 𝑌1 and 𝑌0 are the outcomes for a household in the case of treatment and

no treatment, respectively. D is a dummy variable, indicating that the individual

was treated when it takes the value of 1.

Next, the satisfaction of the conditional mean independence assumption and the

common support condition was verified. As stated by Rosenbaum & Rubin (1983),

the conditional mean independence assumption indicates that after conditioning on

the propensity score, the mean outcomes must be independent from the treatment

state. The common support condition assures that for each treated individual a

potential matched non-treated individual is found, by considering only those

participants whose probability of being treated is lower than 1. Matching can be

done using different matching algorithms that give different weights to the control

units, as described by Caliendo & Kopeinig (2008), the algorithm chosen in this

study was nearest neighbour matching with replacement, with 10 neighbours and

caliper of 0.1, that balanced the trade-off between bias and variance.

After the Propensity Score Matching method controlled for the selection bias on

observables, the Differences-in-differences method was applied to compare the

change in the outcomes of the treated and non-treated for the period of study. The

combination of the Propensity Score Matching with the Differences-in-differences

estimator removes the bias caused by common time trends that are not related to

the treatment, as well as partially overcomes the possibility of selection bias due to

unobserved variables (Heckman et al. 1997). The Differences-in-differences

method compared the observations that were matched using the propensity score,

and was calculated by the equation (5).

25

𝐷𝐼𝐷 =1

𝑁∑[(𝑌𝑖𝑡

1 − 𝑌𝑖𝑡′0 |𝐷 = 1) − ∑ 𝑤𝑖𝑗(𝑌𝑗𝑡−

0 𝑌𝑗𝑡′0 |𝐷 = 0)] (5)

𝐽∩𝑆

𝑗=1

𝐼∩𝑆

𝑖=1

Where N is the number of individuals in the treated group who are in the region

of common support S, i identifies the treated unit (here the household), j identifies

the non-treated unit, t states for the post-treatment period, t’ states for the pre-

treatment period, 𝑤𝑖𝑗 indicates the weights that range between [0, 1], and depend on

the distance between the probability of treatment 𝑃𝑖 and 𝑃𝑗.

Therefore, the average treatment effect on the treated, when combining

Propensity Score Matching with Difference-in-differences, indicated the difference

in the mean growth of the outcome between the group of treated and non-treated.

𝐴𝑇𝑇 = {𝐸(𝑌𝑡1 − 𝑌𝑡′

0|𝐷 = 1, 𝑃(𝑋)) − 𝐸(𝑌𝑡0 − 𝑌𝑡′

0|𝐷 = 0, 𝑃(𝑋))} (6)

26

2.2. The dataset descriptive statistics

The study of the innovations’ impact was performed in two regions of Tanzania,

Morogoro and Dodoma because of their contrasting environmental and socio-

economic conditions that stand out when the regions are compared, and because

both regions represent a majority (between 70 and 80%) of the farming systems

found in the country (Graef et al. 2014). Table 2 presents a comparison of their

characteristics.

Table 2: Summary of Morogoro and Dodoma regions environmental and agricultural

characteristics (Sources: Mnenwa et al. 2010; Graef et al. 2014; World Bank 2020)

2.2.1. Household characteristics

In Dodoma, on average the family size is 5 people and the household head is 50

years old (Table 3). The mean experience in agriculture is 20 years and the average

education is four years. The number of hectares managed by each household are

around 2.5, and in the year 2016, a mean of 80% of was used for cropping. On

average all active members work on-farm, while 40% worked off-farm on the year

2016. The hours worked per hectare increased from year 2013 to year 2016,

reaching about 650. The area perceived as fertile doubled for the same period from

Characteristics Morogoro Dodoma

Biophysical

Climate: semi-humid

Precipitation: 600–800 mm annual

Temperature: average 250C annual.

Varies between 180C to 300C in the

lowlands

Topography: diverse. Flat plains,

highlands, and dry alluvial valleys

Climate: semi-arid

Precipitation: 350–500 mm annual

Temperature: average 220C annual.

Varies between 140C to 300Cds.

Topography: almost homogeneous.

Flat plains and small hills

Socio-economic

Agriculture engages around 70

percent of the region's labour force

Different levels of sensibility

regarding food security

Agriculture engages around 70

percent of the region's labour force

Predominance of high food

insecurity

Agricultural

Major food crops: maize, rice,

sorghum, legumes, horticulture

Major cash crops: sesame,

sunflower, sugarcane, cotton, sisal

Livestock: poultry, cattle, goats.

Secondary source of income

Major food crops: sorghum, millet,

maize

Major cash crops: sesame,

groundnuts, sunflower

Livestock: poultry, cattle, sheep,

goats. Main source of income

27

30 to 60%. While the percentage of farms affected by drought decreased in 20%.

The percentage of households that cultivate cash crops in Dodoma doubled to reach

20% in the year 2016. In the same way, the cultivation of maize doubled in similar

values. The expenditure on fertilizers and pesticides more than doubled between

2013 and 2016 in Dodoma, to reach 8.3 USD per hectare. The household income

perceived from crops increased from 30 to 40%. While the value of the food

expenditures increased to almost 994 USD in the year 2016. The value of

productive assets owned by the households is on average 75 USD.

The average family size in Morogoro is between 4 and 5 persons and the

household head is 50 years old, like Dodoma (Table 4). The average years of

experience in agriculture in Morogoro are less than in Dodoma, though they

increased respect the year 2013 to reach 16 years. The years of education are 4.

Morogoro households manage 2.5 hectares of land from which 80% is cropped area,

same as in Dodoma. The share of cropped area dedicated to cash crops kept constant

in Morogoro, while the share dedicated to maize decreased to 40% in the year 2016.

The labour hours invested per hectare increased to 700 hours in the year 2016.

Similar to Dodoma, in Morogoro on average all active household members work

on-farm while the percentage working off-farm increased to 30%, still lower than

in Dodoma. The share of area perceived as fertile increased to 60%, same level as

Dodoma, though in Morogoro the percentage for year 2013 was higher (50%). On

the other hand, the households affected by drought in Morogoro also increased. The

value of fertilizers and pesticides expenditures per hectare in Morogoro was around

12 USD per hectare, for the years 2013 and 2016, higher than in Dodoma. The

percentage of household income from cropping decreased from 60% to 50% in the

year 2016, contrary to Dodoma where it increased. The value of food expenditures

was around 1055 USD in the year 2016. The value of productive assets doubled

from the year 2013 to the year 2016, reaching almost 64 USD.

28

Variable Unit Dodoma 2013 Dodoma 2016 Difference 2016 - 2013 n mean Sd Min max n mean sd min max

Family size Number

people

448 5 2.2 1 18 420 5 2.2 1 12 0.2

Age of household head Years 447 49 16.9 22 110 420 51 16.9 22 100 2.2

Female household head Yes-No 448 0.2 0.4 0 1 420 0.2 0.4 0 1 0

Experience in agriculture Years 448 20 17.2 0.8 84 420 20 13.6 1 77 0.2

Education of household head Years 424 4 3.4 0 16 415 4 3.5 0 14 0.2

Total labour invested per land unit hours/ha 448 562 748.4 16.4 9452 420 651 508.8 73.3 5647.1 88.9

Share of hired labour in total labour % 448 0.1 0.1 0 1 420 0.1 0.2 0 0.9 0

Active members on-farm % 375 1 0.1 0.2 1 406 1 0.1 0.2 1 0

Active members off-farm % 446 0.4 0.4 0 1 420 0.3 0.4 0 1 -0.1

Total area managed by the household Ha 448 2.4 2.4 0.2 32.1 420 2.5 3.6 0.2 61.3 0.1

Share of total area perceived by the

household as fertile

% 448 0.3 0.4 0 1 420 0.6 0.4 0 1 0.3

Affected by drought Yes-No 448 0.7 0.5 0 1 420 0.5 0.5 0 1 -0.2

Share of total area used for cropping % 448 0.9 0.2 0 1 420 0.8 0.2 0 1 -0.1

Share of cropped area dedicated to

cash crops

% 448 0.1 0.2 0 1 420 0.2 0.2 0 0.8 0.1


maize

% 448 0.1 0.1 0 0.8 420 0.2 0.2 0 1 0.1

Cropped area irrigated Yes-No 448 0 0 0 0.3 420 0 0 0 0.4 0

Value of fertilizer and pesticide

expenditures

USD/ha 448 2.8 11 0 125.8 420 8.3 44.2 0 822.9 5.5

Share of household income generated

by crop production

% 448 0.3 0.3 0 1 420 0.4 0.3 0 1.2 0.1

Value of food expenditures USD 448 855 702.6 0 4288.4 420 993 1153.6 0.3 18421.2 138.5

Value of productive assets USD 448 73 161.6 0 2358.3 420 75 276.2 0 4879.3 2.8

Table 3: Summary statistics of main household characteristics of Dodoma

29

Table 4: Summary statistics of main household characteristics of Morogoro

Variable Unit Morogoro 2013 Morogoro 2016 Difference 2016-2013 n Mean Sd min max n mean sd min max

Family size Number

people

444 4.4 2.3 1 13 391 5 2.4 1 19 0.6

Age of household head Years 443 47 17 19 116 391 50 16 22 100 2.9


Experience in agriculture Years 444 13 11.8 1 82 391 16 12.4 2 87 3.5

Education of household head Years 440 4 3.3 0 14 390 4 3.3 0 17 -0.1

Total labour invested per land unit hours/ha 444 653 737 21 9866.7 389 699 553.1 0 3632.4 46.7

Share of hired labour in total labour % 444 0.2 0.3 0 1 390 0.2 0.3 0 1 0


Active members off-farm % 442 0.2 0.4 0 1 391 0.3 0.4 0 1 0.1



household as fertile

% 444 0.5 0.5 0 1 391 0.6 0.4 0 1 0.1

Affected by drought Yes-No 444 0.1 0.3 0 1 391 0.3 0.5 0 1 0.2

Share of total area used for cropping % 444 0.8 0.2 0 1 391 0.8 0.3 0 1 0

Share of cropped area dedicated to cash

crops

% 444 0.2 0.3 0 1 391 0.2 0.2 0 0.9 0


maize

% 444 0.6 0.3 0 1 391 0.4 0.3 0 1 -0.2

Cropped area irrigated Yes-No 444 0 0.1 0 1 391 0 0.1 0 1 0


expenditures

USD/ha 444 12.5 77.9 0 1028.4 391 12.1 31.2 0 330.7 -0.4

Share of household income generated

by crop production

% 444 0.6 0.3 0 1 391 0.5 0.4 0 1 -0.1

Value of food expenditures USD 444 968 926.4 0 7676.8 390 1054 766.7 0 4298.3 85.7

Value of productive assets USD 444 29 63.1 0 603.2 391 63 331.9 0 6197.1 33.8

30

2.2.2. Sustainability indicators

Disaggregated results about the sustainability indicators’ components in Appendix

1, tables 13 and 14. In the table 5 information about the composite indices for

Dodoma for the years 2013 and 2016 is presented, and the difference between the

years has been calculated.

Table 5: Summary statistics of composite indices for Dodoma

Dodoma 2013

(n = 448)

Dodoma 2016

(n = 420)

Difference 2016 – 2013

Mean Sd mean sd

Soil management 0.36 0.12 0.33 0.11 -0.03***

Water management 0.54 0.12 0.56 0.11 0.02***

Agricultural diversity 0.24 0.14 0.24 0.12 0

Crop performance 0.55 0.09 0.55 0.09 0

Profitability 0.22 0.25 0.28 0.25 0.06***

Stability 0.35 0.34 0.65 0.35 0.3***

Vulnerability reduction 0.7 0.19 0.75 0.15 0.05***

Food security 0.43 0.15 0.54 0.15 0.11***

Health 0.66 0.21 0.65 0.23 -0.01

Wellbeing 0.62 0.24 0.64 0.23 0.02

Social capital 0.24 0.24 0.13 0.13 -0.11***

Land security 0.48 0.19 0.5 0.2 0.02

Environmental index 0.38 0.08 0.38 0.07 0

Economic index 0.46 0.12 0.56 0.13 0.1***

Social index 0.49 0.1 0.49 0.1 0

Overall index 0.44 0.07 0.48 0.08 0.04***

Note: *p<0.1; **p<0.05; ***p<0.01

The overall sustainability index for Dodoma that includes the social, economic

and environmental indicators, increased significantly, possibly related to the rise in

the economic index. There is a significant increase in the economic index and in its

component indicators for profitability, stability, and reduction of vulnerability.

Crop performance had no significant change. The composite environmental index

had no significant change between years, however the indicator for soil

management decreased, while the indicator for water management increased. The

composite social index for Dodoma had no significant change. However, the

indicator for food security increased significantly and the indicator for social capital

decreased significantly.

For Morogoro, the sustainability composite indices for the years 2013 and 2016

and the comparison between years is presented in the table 6.

31

Table 6: Summary statistics of composite indices for Morogoro

Morogoro 2013

(n = 444)

Morogoro 2016

(n = 391)

Difference 2016 – 2013

Mean Sd mean Sd

Soil management 0.3 0.09 0.32 0.1 0.02***

Water management 0.54 0.09 0.54 0.11 0

Agricultural diversity 0.17 0.11 0.25 0.11 0.08***

Crop performance 0.63 0.13 0.54 0.12 -0.09***

Profitability 0.29 0.29 0.32 0.28 0.03

Stability 0.55 0.33 0.57 0.34 0.02

Vulnerability reduction 0.85 0.19 0.73 0.17 -0.12***

Food security 0.53 0.17 0.58 0.14 0.05***

Health 0.63 0.2 0.63 0.23 0

Wellbeing 0.77 0.2 0.6 0.24 -0.17***

Social capital 0.12 0.15 0.1 0.13 -0.02**

Land security 0.39 0.18 0.43 0.2 0.04***

Environmental index 0.34 0.06 0.37 0.07 0.03***

Economic index 0.58 0.15 0.54 0.14 -0.04***

Social index 0.49 0.08 0.47 0.1 -0.02***

Overall index 0.47 0.07 0.46 0.08 -0.01* Note: *p<0.1; **p<0.05; ***p<0.01

The overall sustainability index, that includes the economic, social and

environmental pillars, reduced in Morogoro between 2013 and 2016. The

composite index for environmental sustainability significantly increased. In

contrast to the results for Dodoma, in Morogoro there was a significant increase in

the soil indicator and no change in the water indicator, while the diversity indicator

increased significantly between years. The index for the economic sustainability

had a significant decrease. Respect the indicators that compose it, crop performance

and vulnerability reduction decreased significantly. The composite index for social

sustainability had a significant decrease. There was a significant increase in food

security and land security, but a significant decrease in wellbeing and social capital.

32

3.1. Characteristics associated with the adoption of the

innovations

Before presenting the final results of the Difference-in-differences Propensity Score

Matching method, I present information regarding the covariates used for balancing

in the Propensity Score Matching and the balance achieved. The logit model, as

well as the whole consequent propensity score matching analysis, was carried out

separately for Dodoma and Morogoro due to the differences between their

biophysical characteristics. I chose the covariates expected to influence on the

adoption of the innovations based on their statistical significance and relevance for

the adoption of the treatment according to literature and experts’ opinion. In the

table 7, I present the logistic regressions. For Morogoro 9 variables were chosen

from which 3 are statistically significant, while for Dodoma 7 variables were

chosen from which 2 are statistically significant.

Regarding the balance of the sample, the evaluation through t-tests demonstrated

that only few covariates show a statistically significant difference in the mean

before the matching between the treated and the control group, evaluated at the 10%

level of significance (see Appendix 1 tables 15,16,17,18). In Dodoma, the control

group tended to have a bigger share of cash crops than the treated group. For

Morogoro, adopters tended to have more years of education and a higher land

security indicator than non-adopters. After applying the matching algorithm all the

covariates were balanced for both regions (see Appendix 1 figures 3 and 6). The

variation in results before and after matching shows the usefulness of matching to

reach unbiased results (see unmatched sample results in the Appendix 1 table 19

and 20).

3. Results

33

Table 7: Estimates of the logistic regression for Morogoro and Dodoma

Dependent variable:

Treatment

Morogoro

sd Treatment

Dodoma

sd

Age of household head 0.011 (0.01) -0.004 (0.01)

Education of household head 0.100** (0.05) -0.003 (0.04)

Hired labour -1.142* (0.61)

Total area managed by the household 0.090 (0.06) 0.050 (0.05)

Share of total area used for cropping 0.543 (0.72) 0.800 (0.52)


crops

-0.991 (0.71) -1.328* (0.76)

Share of cropped area dedicated to maize -0.780 (0.63)

Total labour invested per land unit -0.00002 (0.00) -0.00001 (0.00)

Land security index 1.479* (0.75)

Value of productive assets

0.001* (0.00)

Constant -2.964*** (1.04) -1.591** (0.64)

Observations 388 420

Log Likelihood -173.64 -236.72

Akaike Inf. Crit. 367.28 489.45

Note: *p<0.1; **p<0.05; ***p<0.01

The tested balancing property in the covariates allow to use the Difference-in-

differences Propensity Score Matching as an approach to ensure the treated and the

matched control group are comparable and no self-selection biases the outcome

results. The table 8 shows the number of households considered in the Difference-

in-difference regression (sample size after applying the matching algorithm).

34

Table 8: Balanced sample sizes

Number of households Dodoma Morogoro

Treated group

Before matching 110 69

After matching 107 64

Control group

Before matching 310 319

After matching 290 259

Matching algorithm 10:1 nearest neighbour matching with

replacement and caliper of 0.1

3.2. Few impacts of the innovations in Morogoro and

no influence in Dodoma

The estimated impact of adopting rainwater harvesting and fertilizer micro-dosing

over the economic, social and environmental aspects of sustainability of

households-farms of Dodoma and Morogoro regions of Tanzania are presented in

tables 9 and 10. For Morogoro though the composite sustainability was not

impacted by the adoption of the innovations, environmental sustainability and food

security improved, while there was no major effect on the economic and social

sustainability. In Dodoma, the composite sustainability index was not significantly

impacted by the adoption of the innovations, and neither the indicators for each of

the economic, social and environmental pillars of sustainability, though some

components of the sustainability dimensions were impacted to some extent.

35

Table 9: Average Treatment Effect on the Treated (ATT) of Rainwater Harvesting and Fertilizer

Micro-dosing Adoption in Dodoma, 2013–2016 (After matching algorithm applied)

Dependent variable: ATT sd

Mean

Difference of

the Matched

Control Group

Sd

crop gross margin per ha 29.25 (48.54) 27.24 (32.55)

crop expenditures per ha 165.86 (16.44) 15.28* (9.12)

labour productivity 0.11 (0.12) -0.09 (0.08)

post-harvest loss -1.71 (1.34) -0.26 (0.74)

net household income -62.17 (261.58) 1,455.45*** (210.33)

high income fluctuation 0.01 (0.08) -0.02 (0.04)

has savings 0.09* (0.06) 0.61*** (0.04)

loss of income due to shock 111.26 (147.51) 67.57 (59.64)

high severity of shock 0.06 (0.08) -0.06 (0.06)

months to recover after shock -4.52 (6.29) -10.29** (4.26)

Potential Food Availability index

(PFAI) 0.07 (0.3) 0.04 (0.22)

Months of inadequate food

provisioning -0.13 (0.57) -0.96** (0.38)

Food Consumption Score (FCS) 2.80 (2.87) -1.52 (1.93)

Coping Strategies Index (CSI) 1.20 (3.54) -25.94*** (2.64)

quantity of applied fertilizer -378.20 (238.75) 467.07** (206.55)

quantity of animal manure 63.43 (150.85) -108.95 (94.44)

presence of irrigation -0.02 (0.03) 0.01 (0.02)

rainfall water use efficiency 0.13 (0.08) 0.58*** (0.04)

change in household water

consumption -38.49* (20.4) -15.97* (9.63)

water harvesting 0.55*** (0.08) -0.41*** (0.04)

water use conflict 0.07* (0.04) -0.13*** (0.03)

crop performance index 0.01 (0.02) -0.01 (0.01)

profitability index 0.00 (0.02) 0.02** (0.01)

stability index -0.07 (0.07) 0.29*** (0.05)

vulnerability to shocks reduction

index -0.03 (0.03) 0.06*** (0.02)

food security index 0.01 (0.03) 0.09*** (0.02)

soil management index 0.00 (0.02) -0.05*** (0.02)

water management index 0.01 (0.02) 0.03** (0.01)

agricultural diversity index -0.01 (0.03) -0.01 (0.02)

economic sustainability index -0.02 (0.03) 0.09*** (0.02)

social sustainability index 0.02 (0.02) -0.01 (0.01)

environmental sustainability index 0.00 (0.02) -0.01 (0.01)

overall sustainability index 0.00 (0.02) 0.02** (0.01)

Note: *p<0.1; **p<0.05; ***p<0.01

36


Micro-dosing Adoption in Morogoro, 2013–2016 (After matching algorithm applied)

Dependent variable: ATT Sd

Mean

Difference of

the Matched

Control Group

sd

crop gross margin per ha -0.14 (132.67) -216.94** (83.90)

crop expenditures per ha 125.03 (118.89) -67.85 (113.13)

labour productivity 0.24 (0.26) -0.49*** (0.17)

post-harvest loss 0.44 (1.08) 1.28** (0.54)

net household income -81.08 (203.42) 1,360.47*** (144.76)

high income fluctuation -0.04 (0.09) 0.43*** (0.05)

has savings 0.04 (0.09) 0.32*** (0.05)

loss of income due to shock 191.32 (192.69) 656.39*** (71.33)

high severity of shock 0.01 (0.09) 0.46*** (0.05)

months to recover after shock 23.12 (50.72) 18.76 (14.68)

Potential Food Availability index

(PFAI) -0.04 (0.32) 0.19 (0.17)

Months of inadequate food

provisioning -1.01 (0.84) -1.33*** (0.44)

Food Consumption Score (FCS) 5.1* (2.9) -0.33 (1.42)

Coping Strategies Index (CSI) -9.72** (4.53) -7.38*** (1.90)

quantity of applied fertilizer -287.32 (184.22) 288.19 (184.21)

quantity of animal manure 12.50 (10.44) -12.51 (10.44)

presence of irrigation 0.02 (0.03) 0.03 (0.02)

rainfall water use efficiency -0.06 (0.07) 0.5*** (0.05)

change in household water

consumption -28.98* (14.75) 8.75 (9.56)

water harvesting 0.45*** (0.08) -0.04 (0.04)

water use conflict -0.01 (0.04) 0.00 (0.02)

crop performance index 0.03 (0.03) -0.10*** (0.01)

profitability index 0.02 (0.01) -0.02** (0.01)

stability index -0.09 (0.07) 0.01 (0.04)

vulnerability to shocks reduction

index 0.02 (0.04) -0.14*** (0.02)

food security index 0.10*** (0.03) 0.04** (0.02)

soil management index 0.02 (0.02) 0.01 (0.01)

water management index 0.02 (0.02) -0.01 (0.01)

agricultural diversity index 0.04* (0.02) 0.06*** (0.01)

economic sustainability index 0.01 (0.03) -0.06*** (0.02)

social sustainability index 0.02 (0.02) -0.02* (0.01)

environmental sustainability index 0.03** (0.01) 0.02*** (0.01)

overall sustainability index 0.02 (0.01) -0.02** (0.01)

Note: *p<0.1; **p<0.05; ***p<0.01

37

3.2.1. More savings in Dodoma and no impact on economic

sustainability in both regions

In Dodoma, the households that adopted the innovations were more likely to

increase their saving over the period 2013-2016 compared to the non adopters, even

though the composite index for the economic sustainability and its component for

economic stability were not affected by the treatment. Whereas in Morogoro, the

composite index for the economic pillar of sustainability, economic stability and its

components were not impacted significantly. In both regions the adoption of the

innovations had no significant influence in the indicators and composite indices of

crop profitability, profitability, and reduction of vulnerability to shocks.

3.2.2. More environmental sustainability in Morogoro and more

water conflicts in Dodoma

The composite index for environmental sustainability significantly increased by

0.03 by the adoption of the innovations in Morogoro, while there was no impact in

Dodoma. The composite water management index was not significantly impacted

by the adoption of the innovations in Dodoma neither in Morogoro, but in Dodoma

its component for water use conflict increased, and in both regions the change in

household water consumption was impacted. The households that adopted the

innovations in Dodoma reduced less the share of water for which they have

conflicts, between the year 2013 and 2016, by 7% respect to the households that

did not adopt them. In Dodoma, the households that adopted the innovations

reduced the use of water by 38.48 additional daily litres compared to the houses

that did not adopt, while in Morogoro, this reduction was by 28.98 additional daily

litres. The adoption of the innovations had not significant effect on the presence of

irrigation and rainfall use efficiency in both regions, and in water use conflict for

Morogoro.

Respect the composite soil management index, the adoption of the innovations

did not have a significant influence in Dodoma or Morogoro, and neither on the

quantity of animal manure and fertilizer applied. Regarding, the impact on the

agricultural diversity, in Morogoro this indicator had a higher increase between

periods for the households that adopted the innovations, while there was no

significant impact in Dodoma.

38

3.2.3. Improved food security in Morogoro and no impact in

Dodoma

Respect to the food security in Dodoma, the adoption of the innovations rainwater

harvesting and fertilizer micro-dosing did not make any significant difference on

the results of the index or indicators that compose it. Also, the composite index for

the social pillar of sustainability was not affected by the adoption of the innovations.

In Morogoro there was a significant increase on the food security index by the

adoption of the innovations, though the composite index for the social pillar of

sustainability was not impacted. The households that adopted the innovations

increased their food security index by 0.1 additional points in comparison to the

households that did not adopt. The households that adopted the innovations

increased in 5 points more their Food Consumption Score. The Coping Strategies

Index was impacted significantly as well, reducing by 9.7 points more when the

households had adopted the innovations respect to the households that did not. The

Potential Food Availability index was not significantly impacted, and neither the

Months of inadequate food provisioning.

39

The main objective of this research was to estimate the impact of the adoption of

rainwater harvesting and fertilizer micro-dosing over the economic, social and

environmental aspects of sustainability of households-farms of Dodoma and

Morogoro regions of Tanzania. The results show that in the semi-humid region of

Morogoro the adoption of the innovations improved households’ environmental

sustainability indicator by 0.1 and their food security indicator by 0.03, while the

indicators for social sustainability and economic sustainability were not

significantly impacted. In the semi-arid region of Dodoma there is surprisingly no

impacts on households’ composite indices for economic, social or environmental

sustainability and food security, but water conflicts are higher for adopters by 7%,

and they are more probable to have savings than the control group. Adopters of

rainwater harvesting and fertilizer micro-dosing in Morogoro were characterized

by a higher level of education and more land security, while adopters in Dodoma

were characterized by assigning a lower proportion of land to cultivate cash-crops.

4.1. Morogoro: more food security linked to improved

environmental sustainability but no economic

results

In Morogoro, the increase in food security is related to an increase in the quantity

and diversity of food consumed. The adoption of the innovations significantly

improved the yields of maize (see Appendix 1 table 21), the most cultivated crop

in the region, which meets the expectations of Graef et al. (2017) in their expert

based ex-ante analysis for the innovations. The innovations would be tackling the

problem of weather impacts that affected food security, as according to Gornott et

al. (2017) 27% of the loss in maize yields in Tanzania was weather-related.

Furthermore, the innovations appear to be optimizing the use of water in the region,

the environmental sustainability is increased by less household water consumption

and better managed irrigation water: retaining water during high rainfall and

making it available during dry spells. The better environmental conditions

supported by the innovations may also be promoting more diversity in the

4. Discussion

40

agricultural products of Morogoro, as shown by the positive impact in the

agricultural diversity index; this agrees with other cases, like in northern Ethiopia

where rainwater harvesting promoted the cultivation of root crops and vegetables

in previously cereal-based regions (Biazin et al. 2012).

In this research, we could not find evidence that the increase in food security in

Morogoro by the adoption of the innovations is explained by significant economic

benefits, which shows that the innovations are not impacting sustainability in all

the three pillars: economic, social, and environmental, if these aspects are evaluated

in the same period of time. Nevertheless, we cannot assure that the results in this

other pillar may not be achieved after a longer period, the study of van der Marck,

E.J. (1999) indicates that under some conditions rainwater harvesting does not

produce yields that are high enough to justify the investments in labour and

materials, while in the long-run these investments could be covered by the

economic gains and gross margins would be benefitted from the adoption of

rainwater harvesting, (Ellis-Jones & Tengberg 2000). A limitation of this study may

be the comparison between only two periods and the short time frame (only 3

years).

4.2. Dodoma: no food security or economic

improvements motivated by uneven distribution of

water, water conflicts, soil degradation, and

poverty traps

In Dodoma, improvement in household food security with rainwater harvesting and

fertilizer micro-dosing was expected due to its semi-arid climate, the income was

projected to increase and food diversity to improve, according to the ex-ante

assessment done by Graef et al. (2017). However, this could not be demonstrated,

as the results here indicate that adopting rainwater harvesting and fertilizer micro-

dosing had no significant impact in food security and economic sustainability in

Dodoma. Uneven allocation of water, degraded natural resources, and poverty traps

may be influencing on the absence of significant impact. Coping with droughts is

related to food security and reducing poverty, however Enfors & Gordon (2008)

also found lack of effects in stabilizing yields to cope with droughts from the

adoption of another small-scale water system technology in semi-arid Tanzania.

The authors suggested that the reason for the results is the substantial over-use of

the irrigation system, explaining that water allocation among members in water use

groups is uneven and some farmers receive water more frequently. This could be

41

the case for our results if the adopters of rainwater harvesting have also adopted

other water technology innovations. On the other hand, the research of Tittonell &

Giller (2013) about yield gaps that become poverty traps in Africa, states that

cropping continuously without the sufficient nutrients in the soil may degrade this

resource up to the point it becomes non-responsive to fertilizer, driving smallholder

farmers to a poverty trap. More research would be needed to this respect to

determine if Dodoma requires further improvements for the productivity of their

resources. Assets distribution could also generate a poverty trap, Kwak & Smith

(2013) used non parametric techniques to assess the incidence of multiple equilibria

in Ethiopia and found that differences in agricultural assets distribution generate

two equilibria, leaving deprived regions in a low level stagnant equilibrium. The

standard deviation of the value of productive assets shows considerable differences

between farmers in Dodoma, so possibly innovations could work better for some

household types and could have a significant impact.

The absence of impact on the economic indicator could be related to the high

labour demand for applying these innovations and the opportunity costs of labour

in Dodoma. The construction of tied ridges for cultivating under water harvesting

requires 107.6 hours per person per acre, while flat cultivating requires only 24.3

hours per person per acre (Germer et al. 2021). Also, Pender et al. (2006) indicate

that farmers are trying to combine fertilizer micro-dosing with seeds planting to

cope with the labour intensity of fertilizer micro-dosing. In this research, turning to

the household characteristics identified through the Sustainable Livelihoods

Approach, high demand of labour from these practices may be connected to the

increase of labour hours per hectare in Dodoma and Morogoro, however the region

of Dodoma presented a higher percentage of labour working off-farm in the year

2013, which reduced for the year 2016. Therefore, an aspect to evaluate is the

benefits that could be obtained from micro-dosing and water harvesting against the

high labour requirements of the technique, which imply labour opportunity costs

(Sieber et al. 2018; Habtemariam et al. 2019). Also, more economic support may

be needed by farmers under Dodoma’s conditions, Jansen et al. (2006a) agree on

the need for policies prioritizing investments in asset bases for households that have

a relatively high opportunity cost of labour like the ones that have off-farm

employment, as a way to increase returns from assets and raise incomes.

Furthermore, the absence of improvements in economic sustainability of

Dodoma could be indirectly related to the negative results in the indicator for water

conflicts, Nyong (2005) states that the resource degradation motivates proneness to

conflicts, which could be the case of the semi-arid region of Dodoma. Also,

Vohland & Barry (2009) say that rainwater harvesting enhances water infiltration

in the plots, but the reduced downstream runoff may generate conflicts between

42

neighbours who compete for the water resources; conflicts may consume time,

effort and resources.

The results from the study of Läpple & Thorne (2019) that state that a certain

level of innovation must be reached before effects on economic sustainability can

be evidenced, support that for the case of Dodoma it requires to combine the

innovations assessed with others. This approach was proposed by the Trans-SEC

project but here not addressed due to the lower number of adopters of the other

technologies, innovations that may improve food security along with economic

sustainability included adoption of post-harvest processing machines, byproducts

use for bioenergy as well as better market linkages. Promoting policies that

combine higher support for implementation of innovations with training may also

be beneficial, as Jansen et al. (2006a) indicate that training has a direct positive

effect on income improvements.

There was an economic indicator positively affected by the adoption of the

innovations in Dodoma, that stated that adopters were more prompt to have savings

than non-adopters. The explanation of this positive impact may be related to the

perception of farmers that these innovations reduce the risk of water availability,

which as Fox et al. (2005) state rainwater harvesting can reduce the risk of severe

yield reduction over time, from 4 to 1 year out of 10 in their case study for Burkina

Faso and Kenya. The authors also recognize that stability of yields promotes social

resilience, but it is difficult to evaluate in economic terms, such as cost benefit

analysis. Therefore, stability improvements could be better assessed in Dodoma

considering the innovations’ impact for longer periods of time and by a wider set

of measures.

4.3. Benefits from the method: recognizing impacts on

sustainability of more education and land security

in Morogoro and lower cash-crop area in Dodoma

Without the application of any matching algorithm, the mean of the covariates

between treated and non treated group was similar for many of them. However,

some of the covariates are not balanced between the two groups without the

matching and this highlights the importance of matching the groups to avoid biased

conclusions on the effect of the treatment. Indeed, the simple comparison of the

outcomes with the matching and without the matching, shows that some outcomes

are rather different between the two approaches, supporting the decision to

preliminary matching the groups (unmatched sample results in the Appendix 1 table

43

19 and 20). Adopters of the innovations in Dodoma tended to have a lower

proportion of area dedicated to cash crops in comparison to non-adopters, and they

underwent lower economic stability, economic sustainability, and money to invest

in fertilizer and growth of new crops. This goes in accordance to what Maxwell &

Fernando (1989) stated when they analysed the role of cash crops in developing

countries, and they said that although there is a market risk from selling cash crops,

selling a proportion of the output could help to offset the failure of subsistence

crops.

Adopters of the innovations in Morogoro had higher land security and years of

education, which influenced in the significant reduction of their Months of

inadequate food provisioning. More adoption of innovations from farmers with the

mentioned characteristics goes in line with literature, as Jansen et al. (2006a)

concluded that the security in land tenure and education had a significant impact on

the adoption of techniques and innovations. Furthermore, previous studies have also

found a connection between higher levels of education and food security, this is

thought to be related to more adoption of technologies and innovations for

agriculture (Kebede et al. 1990), higher income and better knowledge of nutrition

(De Cock et al. 2013). These findings support the implementation of policies that

facilitate access to land ownership and education in Tanzania, sustained by their

capacity to enhance food security.

The higher number of years of education of the household head and more land

security of adopters in Morogoro may also be related to the reduction in the

economic stability index, since farmers with higher levels of education are more

likely to adopt innovative programmes and grow crops in more competitive markets

for which they expect higher returns. This could be the case of sugarcane production

in Morogoro, which was promoted by the Tanzanian government (Tanzania

National Bureau of Statistics 2017). Morogoro is one of the biggest producers of

sugarcane in Tanzania and in Kilosa farmers produce as outgrowers for the

Kilombero Sugar Company (Chongela 2015), nevertheless the sugarcane

industries tend to delay payments and reduce the price of the good due to cheap

imports of sugar to Tanzania (Smalley et al. 2014).

44

4.4. Limitations of the study

Given the big range of sustainability measurement tools, a comparison to the results

using a different measure could generate a deeper insight about their robustness. As

mentioned earlier, the comparison between only two periods is another limitation,

as the impacts on sustainability may not be evidenced for all pillars of sustainability

after the same period of time. In addition, an analysis that separates household type

could be the next step, which would allow to differentiate the results of the

innovations on different households types under the same biophysical conditions of

one region. This research gives space for new studies comparing the impact on

sustainability of the same innovations but applied under different socio-economic

and biophysical conditions, and measures to overcome poverty traps. Furthermore,

a detailed analysis that explains the impact of the innovations on the decision of

farmers to save and possible effects of the innovations on economic stability over

time could be done.

45

This research supports the use of interdisciplinary assessments along with economic

evaluations in intertemporal analysis, to evaluate the complementarity between

economic, social and environmental sustainability. Assessments that cover the three

pillars of sustainability are recommended to complement the implementation of

agricultural technologies, before and after their use, to recognize the need of

improvement and supplementing actions. Future research for a longer period of

time and possibly considering other methods to make an ex-post sustainability

assessment of these innovations could enrich the debate.

The results of this research highlight the importance of implementing

agricultural technologies, and policies associated to them, targeting at the specific

characteristics and needs of the regions in which the farms are operating. For the

case of semi-humid regions like Morogoro, disseminating the innovations rainwater

harvesting and fertilizer micro-dosing proved to promote environmental

sustainability and ensure food security through more diversity and quantity of food.

Nevertheless, to observe economic benefits from the adoption of the innovations,

the analysis may need to cover a longer period of time. While for the case of regions

like semi-arid Dodoma, it may be worth looking for alternative technologies and

complementing policies, given that no significant impact could be found from the

adoption of rainwater harvesting and fertilizer micro-dosing over social, economic

and environmental sustainability of households and farms. For instance, the

characteristics not balanced according to the Propensity Score Matching method

allowed to recognize that policies promoting cultivation of cash-crops, more

education and land security, can benefit food security and economic results.

Furthermore, counterproductive impacts of agricultural innovations should be

controlled and may be recognized by follow-up assessments to the application of

new technologies, as in this case water conflicts were exacerbated by the adoption

of the innovations in Dodoma.

5. Conclusions

46

Andrade, D., Pasini, F. & Scarano, F.R. (2020). Syntropy and innovation in agriculture. Open Issue 2020 Part A: Technology Innovations and Environmental Sustainability in the Anthropocene, 45, 20–24. https://doi.org/10.1016/j.cosust.2020.08.003

Arata, L. & Sckokai, P. (2016). The Impact of Agri-environmental Schemes on Farm Performance in Five E.U. Member States: A DID-Matching Approach. Land Economics, 92 (1), 167–186

Barrett, C., Reardon, T. & Webb, P. (2001). Nonfarm Income Diversification and Household Livelihood Strategies in Rural Africa: Concepts, Dynamics and Policy Implications. Food Policy, 26, 315–331. https://doi.org/10.1016/S0306-9192(01)00014-8

Biazin, B., Sterk, G., Temesgen, M., Abdulkedir, A. & Stroosnijder, L. (2012). Rainwater harvesting and management in rainfed agricultural systems in sub-Saharan Africa – A review. Recent Advances in Water Resources Management, 47–48, 139–151. https://doi.org/10.1016/j.pce.2011.08.015

Binder, C.R., Feola, G. & Steinberger, J.K. (2010). Considering the normative, systemic and procedural dimensions in indicator-based sustainability assessments in agriculture. Environmental Impact Assessment Review, 30 (2), 71–81. https://doi.org/10.1016/j.eiar.2009.06.002

Bonisoli, L., Galdeano-Gómez, E. & Piedra-Muñoz, L. (2018). Deconstructing criteria and assessment tools to build agri-sustainability indicators and support farmers’ decision-making process. Journal of Cleaner Production, 182, 1080–1094. https://doi.org/10.1016/j.jclepro.2018.02.055

Caliendo, M. & Kopeinig, S. (2008). Some Practical Guidance for the Implementation of Propensity Score Matching. Journal of Economic Surveys, 22 (1), 31–72. https://doi.org/10.1111/j.1467-6419.2007.00527.x

Chabé-Ferret, S. (2015). Analysis of the bias of Matching and Difference-in-Difference under alternative earnings and selection processes. Journal of Econometrics, 185 (1), 110–123. https://doi.org/10.1016/j.jeconom.2014.09.013

Chilagane, E.A., Saidia, P.S., Kahimba, F.C., Asch, F., Germer, J., Graef, F., Swai, E. & Rweyemamu, C.L. (2020). Effects of Fertilizer Micro-dose and In Situ Rain Water Harvesting Technologies on Growth and Yield of Pearl Millet in a Semi-arid Environment. Agricultural Research, 1–13

Chongela, J. (2015). Economic Analysis of Outgrowers’ Sugarcane and Paddy Production Scheme at Ruembe Sugarcane Basin in Kilosa District, Morogoro, Tanzania: A Comparative Approach. Asian Journal of Agricultural Extension, Economics & Sociology, 5, 108–116. https://doi.org/10.9734/AJAEES/2015/14190

Chopin, P., Guindé, L., Causeret, F., Bergkvist, G. & Blazy, J.-M. (2019). Integrating stakeholder preferences into assessment of scenarios for electricity production from locally produced biomass on a small island. Renewable Energy, 131, 128–136. https://doi.org/10.1016/j.renene.2018.06.031

References

47

Clark, N. (2002). Innovation Systems, Institutional Change And The New Knowledge Market: Implications For Third World Agricultural Development. Economics of Innovation and New Technology, 11 (4–5), 353–368. https://doi.org/10.1080/10438590200000004

Comer, S., Ekanem, E., Muhammad, S., Singh, S.P. & Tegegne, F. (1999). Sustainable and Conventional Farmers: A Comparison of Socio-Economic Characteristics, Attitude, and Beliefs. Journal of Sustainable Agriculture, 15 (1), 29–45. https://doi.org/10.1300/J064v15n01_04

De Cock, N., D’Haese, M., Vink, N., van Rooyen, C.J., Staelens, L., Schönfeldt, H.C. & D’Haese, L. (2013). Food security in rural areas of Limpopo province, South Africa. Food Security, 5 (2), 269–282. https://doi.org/10.1007/s12571-013-0247-y

DFID (1999-02-01). DFID sustainable livelihoods guidance sheets Dillon, A. (2011). The Effect of Irrigation on Poverty Reduction, Asset

Accumulation, and Informal Insurance: Evidence from Northern Mali. World Development, 39 (12), 2165–2175. https://doi.org/10.1016/j.worlddev.2011.04.006

Ellis-Jones, J. & Tengberg, A. (2000). The impact of indigenous soil and water conservation practices on soil productivity: examples from Kenya, Tanzania and Uganda. Land Degradation & Development, 11 (1), 19–36. https://doi.org/10.1002/(SICI)1099-145X(200001/02)11:1<19::AID-LDR357>3.0.CO;2-2

Enfors, E.I. & Gordon, L.J. (2008). Dealing with drought: The challenge of using water system technologies to break dryland poverty traps. Local evidence on vulnerabilities and adaptations to global environmental change, 18 (4), 607–616. https://doi.org/10.1016/j.gloenvcha.2008.07.006

Food and Agriculture Organization (2021). The World Bank. https://data.worldbank.org/indicator/SN.ITK.DEFC.ZS?locations=ZG-TZ [2021-04-19]

Fox, P., Rockström, J. & Barron, J. (2005). Risk analysis and economic viability of water harvesting for supplemental irrigation in semi-arid Burkina Faso and Kenya. Agricultural Systems, 83 (3), 231–250. https://doi.org/10.1016/j.agsy.2004.04.002

Friedrich List (1841). The National System of Political Economy. London: Longman.

Germer, J., Herrman, L., Mahoo, H., Swai, E., Graef, F., Kahimba, F., Asch, F., Tumbo, S., Makoro, B., Kimaro, A., Schäfer, M., Saidia, P. & Chilagane, E. (2021). Trans-SEC Factsheet 1a. Trans-SEC. http://project2.zalf.de/trans-sec/public/media/factsheets/Trans-SECfactsheet1a.pdf

Gómez-Limón, J.A. & Sanchez-Fernandez, G. (2010). Empirical evaluation of agricultural sustainability using composite indicators. Ecological Economics, 69 (5), 1062–1075. https://doi.org/10.1016/j.ecolecon.2009.11.027

Gornott, C., Hattermann, F., Nkonya, E., Herrmann, L., Graef, F. & Tumbo, S.K. (2017). Trans-SEC scenario definition and framework.

Graef, F., Sieber, S., Mutabazi, K., Asch, F., Biesalski, H., Bitegeko, J., Bokelmann, W., Brüntrup, M., Dietrich, O., Elly, N., Faße, A., Germer, J., Grote, U., Herrmann, L., Herrmann, R., Hoffmann, H., Kahimba, F., Kaufmann, B., Kersebaum, K. & Uckert, G. (2014). Framework for participatory food security research in rural food value chains. Global Food Security, Vol. 3, 8–15. https://doi.org/10.1016/j.gfs.2014.01.001

Graef, F., Uckert, G., Schindler, J., König, H.J., Mbwana, H.A., Fasse, A., Mwinuka, L., Mahoo, H., Kaburire, L.N., Saidia, P., Yustas, Y.M., Silayo, V., Makoko, B., Kissoly, L., Lambert, C., Kimaro, A., Sieber, S., Hoffmann,

48

H., Kahimba, F.C. & Mutabazi, K.D. (2017). Expert-based ex-ante assessments of potential social, ecological, and economic impacts of upgrading strategies for improving food security in rural Tanzania using the ScalA-FS approach. Food Security, 9 (6), 1255–1270. https://doi.org/10.1007/s12571-016-0639-x

Habtemariam, L.T., Mgeni, C.P., Mutabazi, K.D. & Sieber, S. (2019). The farm income and food security implications of adopting fertilizer micro-dosing and tied-ridge technologies under semi-arid environments in central Tanzania. Journal of Arid Environments, 166, 60–67. https://doi.org/10.1016/j.jaridenv.2019.02.011

Haileslassie, A., Craufurd, P., Thiagarajah, R., Kumar, S., Whitbread, A., Rathor, A., Blummel, M., Ericsson, P. & Kakumanu, K.R. (2016). Empirical evaluation of sustainability of divergent farms in the dryland farming systems of India. Ecological Indicators, 60, 710–723. https://doi.org/10.1016/j.ecolind.2015.08.014

ul Haq, S. & Boz, I. (2018). Developing a set of indicators to measure sustainability of tea cultivating farms in Rize Province, Turkey. Ecological Indicators, 95, 219–232. https://doi.org/10.1016/j.ecolind.2018.07.041

Harvey, C.A., Rakotobe, Z.L., Rao, N.S., Dave, R., Razafimahatratra, H., Rabarijohn, R.H., Rajaofara, H. & MacKinnon, J.L. (2014). Extreme vulnerability of smallholder farmers to agricultural risks and climate change in Madagascar. Philosophical Transactions of the Royal Society B: Biological Sciences, 369 (1639), 20130089. https://doi.org/10.1098/rstb.2013.0089

Hatibu, N., Mutabazi, K., Senkondo, E.M. & Msangi, A.S.K. (2006). Economics of rainwater harvesting for crop enterprises in semi-arid areas of East Africa. Special Issue on Water Scarcity: Challenges and Opportunities for Crop Science, 80 (1), 74–86. https://doi.org/10.1016/j.agwat.2005.07.005

ICRISAT (2021). Microdosing – a key proven technology for sub-Saharan Africa – ICRISAT. https://www.icrisat.org/microdosing-a-key-proven-technology-for-sub-saharan-africa/ [2021-05-15]

Jansen, H.G.P., Pender, J., Damon, A., Wielemaker, W. & Schipper, R. (2006a). Policies for sustainable development in the hillside areas of Honduras: a quantitative livelihoods approach. Agricultural Economics, 34 (2), 141–153. https://doi.org/10.1111/j.1574-0864.2006.00114.x

Jansen, H.G.P., Rodriguez, A., Damon, A., Pender, J., Chenier, J. & Schipper, R. (2006b). Determinants of income-earning strategies and adoption of conservation practices in hillside communities in rural Honduras. Heterogeneity and Diversity in Less-Favoured Areas, 88 (1), 92–110. https://doi.org/10.1016/j.agsy.2005.06.005

Kebede, Y., Gunjal, K. & Coffin, G. (1990). Adoption of New Technologies in Ethiopian Agriculture: The Case of Tegulet-Bulga District, Shoa Province. Agricultural Economics, 4 (1), 27–43. https://doi.org/10.1111/j.1574-0862.1990.tb00103.x

Kwak, S. & Smith, S.C. (2013). Regional Agricultural Endowments and Shifts of Poverty Trap Equilibria: Evidence from Ethiopian Panel Data. The Journal of Development Studies, 49 (7), 955–975. https://doi.org/10.1080/00220388.2013.785523

Lana, M.A., Vasconcelos, A.C.F., Gornott, C., Schaffert, A., Bonatti, M., Volk, J., Graef, F., Kersebaum, K.C. & Sieber, S. (2018). Is dry soil planting an adaptation strategy for maize cultivation in semi-arid Tanzania? Food Security, 10 (4), 897–910. https://doi.org/10.1007/s12571-017-0742-7

Läpple, D. & Thorne, F. (2019). The Role of Innovation in Farm Economic Sustainability: Generalised Propensity Score Evidence from Irish Dairy

49

Farms. Journal of Agricultural Economics, 70 (1), 178–197. https://doi.org/10.1111/1477-9552.12282

Lee, D.R. (2005). Agricultural Sustainability and Technology Adoption: Issues and Policies for Developing Countries. American Journal of Agricultural Economics, 87 (5), 1325–1334. https://doi.org/10.1111/j.1467-8276.2005.00826.x

Maxwell, S. & Fernando, A. (1989). Cash crops in developing countries: The issues, the facts, the policies. World Development, 17 (11), 1677–1708. https://doi.org/10.1016/0305-750X(89)90193-9

Mmbaga, T.E., Friesen, D.K. |Palmer, A.F.E. & Lyamchai, C.Y. (2002). Drought management options in maize production in northern Tanzania. (Centro Internacional de Mejoramiento de Maiz y Trigo (CIMMYT) Kenya & 7. Proceedings of the Eastern and Southern Africa Regional Maize Conference, eds.). Nairobi (Kenya): KARI|CIMMYT.

Mnenwa, R., Maliti, E., & Research on Poverty Alleviation (Tanzania) (2010). A comparative analysis of poverty incidence in farming systems of Tanzania. Dar es Salaam, Tanzania: Research on Poverty Alleviation.

Mutyasira, V., Hoag, D., Pendell, D., Manning, D.T. & Berhe, M. (2018). Assessing the relative sustainability of smallholder farming systems in Ethiopian highlands. Agricultural Systems, 167, 83–91. https://doi.org/10.1016/j.agsy.2018.08.006

Nyong, A. (2005). A message to the G8 summit. Nature 435. de Olde, E.M., Oudshoorn, F.W., Sørensen, C.A.G., Bokkers, E.A.M. & de Boer,

I.J.M. (2016). Assessing sustainability at farm-level: Lessons learned from a comparison of tools in practice. Ecological Indicators, 66, 391–404. https://doi.org/10.1016/j.ecolind.2016.01.047

Pender, J., Abdoulaye, T., Ndjeunga, J., Gerard, B. & Edward, K. (2006). Impacts of Inventory Credit, Input Supply Shops and Fertilizer Micro-Dosing in the Drylands of Niger.

Pender, J., Scherr, S. & Durón, G. (2001). Pathways of development in the hillsides of Honduras: causes and implications for agricultural production, poverty, and sustainable resource use. Proceedings of In: Lee, D. R., Barrett, C. B. (Eds.), Tradeoffs or synergies? Agricultural intensification, economic development and the environment, Wallingford, UK., 2001. Wallingford, UK.: CAB International

Quandt, A. (2018). Measuring livelihood resilience: The Household Livelihood Resilience Approach (HLRA). World Development, 107, 253–263. https://doi.org/10.1016/j.worlddev.2018.02.024

Rakodi, C. (1999). A Capital Assets Framework for Analysing Household Livelihood Strategies: Implications for Policy. Development Policy Review, 17 (3), 315–342. https://doi.org/10.1111/1467-7679.00090

Romer, P.M. (1986). Increasing Returns and Long-Run Growth. Journal of Political Economy, 94 (5), 1002–1037

Rosenbaum, P.R. & Rubin, D.B. (1983). The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika, 70 (1), 41–55. https://doi.org/10.2307/2335942

Saidia, P.S., Asch, F., Kimaro, A.A., Germer, J., Kahimba, F.C., Graef, F., Semoka, J.M.R. & Rweyemamu, C.L. (2019). Soil moisture management and fertilizer micro-dosing on yield and land utilization efficiency of inter-cropping maize-pigeon-pea in sub humid Tanzania. Agricultural Water Management, 223, 105712. https://doi.org/10.1016/j.agwat.2019.105712

Schindler, J., Graef, F. & König, H.J. (2016a). Participatory impact assessment: Bridging the gap between scientists’ theory and farmers’ practice. Agricultural Systems, 148, 38–43. https://doi.org/10.1016/j.agsy.2016.07.002

50

Schindler, J., Graef, F., König, H.J., Mchau, D., Saidia, P. & Sieber, S. (2016b). Sustainability impact assessment to improve food security of smallholders in Tanzania. Environmental Impact Assessment Review, 60, 52–63. https://doi.org/10.1016/j.eiar.2016.04.006

Sieber, S., Graef, F., Amjath-Babu, T.S., Mutabazi, K.D., Tumbo, S.D., Faße, A., Gomez y Paloma, S., Rybak, C., Lana, M.A., Ndah, H.T., Uckert, G., Schuler, J. & Grote, U. (2018). Trans-SEC’s food security research in Tanzania: from constraints to adoption for out- and upscaling of agricultural innovations. Food Security, 10 (4), 775–783. https://doi.org/10.1007/s12571-018-0822-3

Smalley, R., Sulle, E. & Malale, L. (2014). FAC Working Paper 106. The role of the state and foreign capital in agricultural commercialisation:The case of sugarcane outgrowers in Kilombero District, Tanzania., 2014.

Sumberg, J. (2005). Systems of innovation theory and the changing architecture of agricultural research in Africa. Food Policy, 30 (1), 21–41. https://doi.org/10.1016/j.foodpol.2004.11.001

Tanzania National Bureau of Statistics (2017). National Bureau of Statistics - Home. https://www.nbs.go.tz/index.php/en/ [2021-05-14]

The World Bank (1991). The challenge of development. Oxford: Oxford Univ. Press. (World development report; [14.]1991)

The World Bank (2020-10-13). Tanzania Overview. https://www.worldbank.org/en/country/tanzania/overview#1 [2021-03-16]

Tittonell, P. & Giller, K.E. (2013). When yield gaps are poverty traps: The paradigm of ecological intensification in African smallholder agriculture. Crop Yield Gap Analysis – Rationale, Methods and Applications, 143, 76–90. https://doi.org/10.1016/j.fcr.2012.10.007

Udagawa, C., Hodge, I. & Reader, M. (2014). Farm Level Costs of Agri-environment Measures: The Impact of Entry Level Stewardship on Cereal Farm Incomes. Journal of Agricultural Economics, 65 (1), 212–233. https://doi.org/10.1111/1477-9552.12043

UNICEF (2020-07-13). The State of Food Security and Nutrition in the World 2020. UNICEF DATA. https://data.unicef.org/resources/sofi-2020/ [2021-04-19]

United Nations Development Programme (2015). Goal 2: Zero hunger. UNDP. https://www.undp.org/content/undp/en/home/sustainable-development-goals/goal-2-zero-hunger.html [2021-04-12]

United Nations Framework Convention on Climate Change (2021-03-18). Climate Change-Related Disasters a Major Threat to Food Security - FAO | UNFCCC. https://unfccc.int/news/climate-change-related-disasters-a-major-threat-to-food-security-fao [2021-04-19]

van der Marck, E.J. (1999). A Model for the Optimization of Water Harvesting Schemes. Enschede: University of Twente.

Vohland, K. & Barry, B. (2009). A review of in situ rainwater harvesting (RWH) practices modifying landscape functions in African drylands. Agriculture, Ecosystems & Environment, 131 (3), 119–127. https://doi.org/10.1016/j.agee.2009.01.010

World Bank (2020). Supporting Countries in Unprecedented Times Annual Report 2020

World Food Programme (2021). Tanzania | World Food Programme. https://www.wfp.org/countries/tanzania [2021-04-12]

Yegbemey, R.N., Yabi, J.A., Dossa, C.S.G. & Bauer, S. (2014). Novel participatory indicators of sustainability reveal weaknesses of maize cropping in Benin. Agronomy for Sustainable Development, 34 (4), 909–920. https://doi.org/10.1007/s13593-014-0214-9

51

My special thanks to my parents who have been my motivation and support to

pursue these studies, and to my brothers and grandfather that are always looking

after me.

I would like to express my gratitude to Pierre Chopin, for his time and guidance

that have been key to develop this work and improve my academic skills, and to

Jens Rommel, Paolo Sckokai, Linda Arata and Kai Mausch for all the valuable

contributions and advise.

Acknowledgements

52

Table 11: Summary statistics of household characteristics of Dodoma

Dodoma 2013 Dodoma 2016 Difference

2016 - 2013

Variable Units n mean sd min max n mean sd min max

Family size No.

people 448 5.3 2.2 1 18 420 5.5 2.2 1 12 0.2

Age of household head Years 447 49 16.9 22 110 420 51.2 16.9 22 100 2.2


Experience of household head Years 448 20.5 17.2 0.8 84 420 20.7 13.6 1 77 0.2

Education of household head Years 424 4.2 3.4 0 16 415 4.4 3.5 0 14 0.2

Household female literacy % 436 0.5 0.5 0 1 407 0.6 0.4 0 1 0.1

Total labour invested per land unit hours/ha 448 562.4 748.4 16.4 9452 420 651.3 508.8 73.3 5647.1 88.9

Share of hired labour in total labour % 448 0.1 0.1 0 1 420 0.1 0.2 0 0.9 0

Share of total labour working in land

preparation %

448 0.2 0.2 0 0.8 420 0.2 0.1 0 0.8 0

Share of total labour working in weeding % 448 0.3 0.2 0 1.1 420 0.3 0.1 0 0.8 0

Share of total labour working in

harvesting %

448 0.2 0.2 0 2.2 420 0.2 0.1 0 1 0

Share of total labour working in planting % 448 0.1 0.1 0 0.6 420 0.1 0.1 0 0.4 0

Share of total labour working in other

cropping activities %

448 0.1 0.1 0 0.8 420 0.1 0.1 0 0.3 0

Appendix 1

53


Active members off-farm % 446 0.4 0.4 0 1 420 0.3 0.4 0 1 -0.1


Share of total area rented in % 448 0.1 0.2 0 1 420 0.1 0.2 0 1 0

Share of crop land managed remotely % 442 0.4 0.4 0 1 409 0.3 0.4 0 1 -0.1


household as fertile %

448 0.3 0.4 0 1 420 0.6 0.4 0 1 0.3

Affected by drought Yes-No 448 0.7 0.5 0 1 420 0.5 0.5 0 1 -0.2

Share of total area used for cropping % 448 0.9 0.2 0 1 420 0.8 0.2 0 1 -0.1


grassland %

448 0 0 0 0.9 420 0 0 0 0.8 0


crops %

448 0.1 0.2 0 1 420 0.2 0.2 0 0.8 0.1

Share of cropped area dedicated to maize % 448 0.1 0.1 0 0.8 420 0.2 0.2 0 1 0.1


bullrush %

448 0.3 0.3 0 1 420 0.3 0.3 0 1 0

Share of cropped area dedicated to other

cereals %

448 0.2 0.3 0 1 420 0.2 0.3 0 1 0


legumes %

448 0.3 0.2 0 1 420 0.2 0.2 0 1 -0.1


crops %

448 0 0.1 0 0.6 420 0 0.1 0 1 0


cereals %

448 0.6 0.2 0 1 420 0.6 0.2 0 1 0

Cropped area irrigated Yes-No 448 0 0 0 0.3 420 0 0 0 0.4 0


expenditures USD/ha

448 2.8 11 0 125.8 420 8.3 44.2 0 822.9 5.5

Household Tropical Livestock Units

(TLU) TLU

448 1.4 2.7 0 21 420 1.3 2.8 0 28.7 -0.1

Size of poultry herd TLU 448 0 0.1 0 0.6 420 0.1 0.1 0 1.2 0.1

54

Size of goat and sheep herd TLU 448 0.2 0.7 0 8 420 0.2 0.4 0 5 0

Size of pig herd TLU 448 0.1 0.2 0 2.2 420 0.1 0.3 0 2.4 0

Size of cattle herd TLU 448 1 2.3 0 18.9 420 1 2.5 0 27.3 0

Share of crop production sold % 448 0.1 0.2 0 1 420 0.2 0.2 0 1 0.1

Share of household income generated by

crop production %

448 0.3 0.3 0 1 420 0.4 0.3 0 1.2 0.1


livestock production %

448 0.2 0.3 0 1 420 0.2 0.2 0 1 0


off-farm employment %

448 0.1 0.2 0 1 420 0.1 0.2 0 1 0


self-employment %

448 0.1 0.2 0 1 420 0.2 0.3 0 1 0.1


hunting/gathering %

448 0.2 0.3 0 1 420 0 0.1 -0.2 1 -0.2

Share of household income from

remittances %

448 0 0.1 0 0.7 420 0.1 0.2 0 1 0.1

Share of household income from social

support systems %

448 0 0.1 0 0.8 420 0 0.1 0 1 0


off-farm activities %

448 0.5 0.3 0 1 420 0.4 0.3 -0.2 1 -0.1

Value of loans USD 448 33.5 163.2 0 2545.4 420 41.8 308 0 5904.2 8.3

Value of household expenditures USD 448 1432.7 1072.8 0 5790.2 420 1496.4 1418 93.9

19315.

1 63.7

Value of food expenditures USD 448 855.2 702.6 0 4288.4 420 993.7 1153.6 0.3

18421.

2 138.5

Number of ploughs No. 448 0.3 0.6 0 5 420 0.3 0.6 0 3 0

Number of stoves No. 448 0.1 0.3 0 2 420 0.1 0.4 0 4 0

Number of mobile phones No. 448 0.5 0.8 0 4 420 0.8 0.9 0 5 0.3

Value of productive assets USD 448 73.1 161.6 0 2358.3 420 75.9 276.2 0 4879.3 2.8

Value of household assets USD 448 153.8 339.3 0 3347.1 420 219.2 449.6 0 3457.5 65.4

55

Table 12: Summary statistics of household characteristics of Morogoro

Morogoro 2013 Morogoro 2016 Difference

2016 - 2013

Variable Units n mean sd min max n mean sd min max

Family size No.

people 444 4.4 2.3 1 13 391 5 2.4 1 19 0.6

Age of household head Years 443 47.9 17 19 116 391 50.8 16 22 100 2.9


Experience of household head Years 444 13.3 11.8 1 82 391 16.8 12.4 2 87 3.5

Education of household head Years 440 4.9 3.3 0 14 390 4.8 3.3 0 17 -0.1

Household female literacy % 409 0.7 0.4 0 1 372 0.6 0.4 0 1 -0.1

Total labour invested per land unit hours/ha 444 653.2 737 21 9866.7 389 699.9 553.1 0 3632.4 46.7

Share of hired labour in total labour % 444 0.2 0.3 0 1 390 0.2 0.3 0 1 0

Share of total labour working in land

preparation %

444 0.3 0.2 0 0.8 390 0.2 0.2 0 0.8 -0.1

Share of total labour working in weeding % 444 0.3 0.2 0 2.1 390 0.3 0.2 0 1.3 0

Share of total labour working in

harvesting %

444 0.2 0.1 0 0.9 390 0.2 0.2 0 1.1 0

Share of total labour working in planting % 444 0.1 0.1 0 0.8 390 0.1 0.1 0 0.7 0

Share of total labour working in other

cropping activities %

444 0.1 0.1 0 0.8 390 0.1 0.1 0 0.5 0


Active members off-farm % 442 0.2 0.4 0 1 391 0.3 0.4 0 1 0.1


Share of total area rented in % 444 0.2 0.3 0 1 391 0.1 0.3 0 1 -0.1

Share of crop land managed remotely % 430 0.3 0.4 0 1 380 0.2 0.4 0 1 -0.1


household as fertile %

444 0.5 0.5 0 1 391 0.6 0.4 0 1 0.1

Affected by drought Yes-No 444 0.1 0.3 0 1 391 0.3 0.5 0 1 0.2

56

Share of total area used for cropping % 444 0.8 0.2 0 1 391 0.8 0.3 0 1 0


grassland %

444 0 0.1 0 1 391 0 0 0 0 0


crops %

444 0.2 0.3 0 1 391 0.2 0.2 0 0.9 0

Share of cropped area dedicated to maize % 444 0.6 0.3 0 1 391 0.4 0.3 0 1 -0.2


bullrush %

444 0 0 0 0.2 391 0.1 0.2 0 1 0.1


cereals %

444 0.1 0.2 0 1 391 0.1 0.2 0 1 0


legumes %

444 0.1 0.1 0 0.8 391 0.2 0.2 0 0.8 0.1


crops %

444 0 0.1 0 1 391 0 0.1 0 0.5 0


cereals %

444 0.7 0.3 0 1 391 0.6 0.2 0 1 -0.1

Cropped area irrigated Yes-No 444 0 0.1 0 1 391 0 0.1 0 1 0


expenditures USD/ha

444 12.5 77.9 0 1028.4 391 12.1 31.2 0 330.7 -0.4

Household Tropical Livestock Units

(TLU) TLU

444 0.3 1.7 0 21.6 391 0.7 2.5 0 40.2 0.4

Size of poultry herd TLU 444 0.1 0.2 0 2.2 391 0.1 0.1 0 0.9 0

Size of goat and sheep herd TLU 444 0.1 0.5 0 10 391 0.1 0.8 0 11 0

Size of pig herd TLU 444 0 0.1 0 2 391 0 0.2 0 2.4 0

Size of cattle herd TLU 444 0.2 1.6 0 19.6 391 0.4 1.9 0 29.4 0.2

Share of crop production sold % 444 0.4 0.3 0 1 391 0.4 0.3 0 1 0


crop production %

444 0.6 0.3 0 1 391 0.5 0.4 0 1 -0.1


livestock production %

444 0.1 0.2 0 1 391 0.1 0.2 0 1 0

57


off-farm employment %

444 0 0.1 0 0.9 391 0.1 0.2 0 1 0.1


self-employment %

444 0.1 0.2 0 1 391 0.2 0.3 0 1 0.1


hunting/gathering %

444 0.1 0.2 0 1 391 0 0.1 0 1 -0.1

Share of household income from

remittances %

444 0 0.1 0 1 391 0.1 0.2 0 1 0.1

Share of household income from social

support systems %

444 0 0 0 0.7 391 0 0.1 0 1 0


off-farm activities %

444 0.3 0.3 0 1 391 0.4 0.4 0 1 0.1

Value of loans USD 444 50.1 466.2 0 8908.7 391 50.7 206.5 0 2361.7 0.6

Value of household expenditures USD 444 1637 1539.5 0

11153.

7 390 1618.6 1127.9 29.5 8549.3 -18.4

Value of food expenditures USD 444 968.9 926.4 0 7676.8 390 1054.6 766.7 0 4298.3 85.7

Number of ploughs No. 444 0 0.1 0 1 391 0.1 0.3 0 3 0.1

Number of stoves No. 444 0.2 0.4 0 4 391 0.2 0.5 0 2 0

Number of mobile phones No. 444 0.7 0.8 0 5 391 0.9 0.8 0 4 0.2

Value of productive assets USD 444 29.9 63.1 0 603.2 391 63.7 331.9 0 6197.1 33.8

Value of household assets USD 444 215.5 444.3 0 3958 391 320.7 622.3 0 5845.2 105.2

58

Table 13: Summary statistics of sustainability indicators for Dodoma

Dodoma 2013

(n = 448)

Dodoma 2016

(n = 420)

Difference 2016 - 2013 Pillar Component Indicator Mean

CV

(%) mean CV (%)

Envir

onm

enta

l

Soil

management

Quantity of bought

fertilizer 157.7 484 611.8 740 454.1

Quantity of animal

manure 135.8 580 82.4 592 -53.4

Crop area perceived

as unfertile 0.2 200 0 0 -0.2

Crop area with

perceived decrease in

fertility 0.6 67 0.5 80 -0.1

Crop area under

legume 0.3 67 0.2 100 -0.1

Crop residues left on

the field 0.3 133 0.2 150 -0.1

Intent to invest in soil

fertility 0.8 50 0.5 80 -0.3

Tree density 11.4 532 9.4 400 -2

Area under erosion

control measures 0.3 133 0.2 200 -0.1

Water

management

Presence of irrigation 0 0 0 0 0

Rainfall water use

efficiency 0.5 200 0.6 117 0.1

Change in water

consumption 7.1 549 -20.9 -642 -28

Water harvesting 0.4 125 0.2 200 -0.2

Water use conflict 0.1 300 0 0 -0.1

Agricultural

diversity

Tree diversity 2.9 83 2.8 64 -0.1

Crop diversity 3.7 43 3.3 39 -0.4

Livestock diversity 1.3 100 1.8 72 0.5

Eco

no

mic

Crop

profitability

Crop gross margin 159 229 206.4 129 47.4

Crop expenditures 35.8 206 53.4 171 17.6

Labour productivity 0.6 217 0.5 100 -0.1

Post-harvest loss 3.1 319 2.1 205 -1

Profitability Net household

income 982.5 133 1368.8 126 386.3

Stability

High income

fluctuation 0.5 100 0.5 100 0

Has savings 0.2 200 0.8 50 0.6

Vulnerability

to shocks

Loss of income due

to shock 349.6 202 424.4 161 74.8

High severity of

shock 0.7 57 0.7 71 0


shock 25.6 130 18.5 588 -7.1

59

So

cial

Food security

Potential Food

Availability index

(PFAI) 2 75 2 140 0

Months of

inadequate food

provisioning 5.8 62 4.8 77 -1

Food Consumption

Score (FCS) 45.8 35 46.9 38 1.1

Coping Strategies

Index (CSI) 33.8 88 6.3 244 -27.5

Health

Health insurance

binary 0.3 133 0.3 167 0

Healthy household

members 0.8 37 0.8 25 0

Wellbeing

Hours worked 664.3 120 707.2 97 42.9

Perceived

deterioration of

household situation 0.5 100 0.4 125 -0.1

High impact of

income fluctuations

on wellbeing 0.5 100 0.4 125 -0.1

Social capital

Information network 34.2 142 14.3 99 -19.9



group 0.1 300 0.1 200 0

Land security

Land title ownership 0 0 0 0 0

Secure land 0.5 100 0.5 100 0

Land use conflict 0.1 200 0 0 -0.1

60

Table 14: Summary statistics of sustainability indicators for Morogoro

Morogoro 2013

(n = 444)

Morogoro 2016

(n = 391) Difference 2016 - 2013 Pillar Component Indicator mean CV (%) mean CV (%)

En

vir

on

men

tal

Soil

management

Quantity of bought

fertilizer 2 930 298.4 1424 296.4

Quantity of animal

manure 26 1316 0.5 1240 -25.5

Crop area perceived as

unfertile 0.1 200 0 0 -0.1

Crop area with

perceived decrease in

fertility 0.5 100 0.3 133 -0.2

Crop area under

legume 0.1 100 0.2 100 0.1

Crop residues left on

the field 0.4 125 0.1 200 -0.3

Intent to invest in soil

fertility 0.3 167 0.3 133 0

Tree density 4.6 165 7.9 194 3.3

Area under erosion

control measures 0.2 150 0.2 150 0

Water

management

Presence of irrigation 0 0 0.1 300 0.1

Rainfall water use

efficiency 0.7 243 0.5 160 -0.2

Change in water

consumption 1.1 3573 -2.5 -3328 -3.6

Water harvesting 0.1 300 0.2 200 0.1

Water use conflict 0.1 200 0 0 -0.1

Agricultural

diversity

Tree diversity 1.8 100 2.4 67 0.6

Crop diversity 2.1 48 2.8 39 0.7

Livestock diversity 0.8 87 0.9 67 0.1

Eco

no

mic

Crop

profitability

Crop gross margin 609.3 175 391 210 -218.3

Crop expenditures 195.8 589 191.8 173 -4

Labour productivity 1.6 125 1.1 173 -0.5

Post-harvest loss 1.8 322 3.5 249 1.7

Profitability Net household income 1483.9 187 1396.2 117 -87.7

Stability

High income

fluctuation 0.3 133 0.6 83 0.3

Has savings 0.4 125 0.7 71 0.3

Vulnerability

to shocks

Loss of income due to

shock 155.8 274 768.2 136 612.4

High severity of shock 0.4 125 0.8 50 0.4


shock 11.9 223 48 691 36.1

61

So

cial

Food

security

Potential Food

Availability index

(PFAI) 2.7 78 2.7 74 0

Months of inadequate

food provisioning 5.1 96 4 88 -1.1

Food Consumption

Score (FCS) 53.2 31 53.8 29 0.6

Coping Strategies

Index (CSI) 16.2 149 7.3 225 -8.9

Health

Health insurance

binary 0.1 300 0.2 200 0.1

Healthy household

members 0.9 33 0.8 37 -0.1

Wellbeing

Hours worked 532.1 154 627 124 94.9

Perceived deterioration

of household situation 0.3 167 0.6 83 0.3

High impact of income

fluctuations on

wellbeing 0.2 200 0.5 100 0.3

Social

capital

Information network 16.3 133 28.1 142 11.8



group 0.3 133 0.1 300 -0.2

Land

security

Land title ownership 0.1 300 0 0 -0.1

Secure land 0.1 300 0.3 133 0.2

Land use conflict 0.1 300 0.1 300 0

62

Table 15: Summary of Balance after matching in Dodoma

Means

Treated

Means

Control

Std. Mean

Diff.

Var.

Ratio

p-value

distance 0.2707 0.2706 0.0017 1.0129

Age of

household head 49.4746 49.5673 -0.0061 0.7636

0.79

Education of

household head 4.2461 4.2270 0.0061 0.8023

0.62

Total labour

invested per land

unit

559.2268 506.6087 0.0738 1.6210

0.88

Total area

managed by the

household

2.5440 3.4118 -0.4118 0.0971

0.20

Share of total

area used for

cropping

0.8861 0.8570 0.1497 0.5967

0.22

Share of cropped

area dedicated to

cash crops

0.0958 0.1084 -0.0981 0.6692

0.10

Value of

productive assets 74.5890 71.2062 0.0137 0.5643

0.22

Table 16: Summary of Balance before matching in Dodoma

Means

Treated

Means

Control

Std. Mean

Diff.

Var.

Ratio

p-value

distance 0.2786 0.2560 1.6411

Age of

household head

49.5253 49.6839 -0.0104 0.7979 0.93

Education of

household head

4.2485 4.1363 0.0359 0.8491 0.75

Total labour

invested per land

unit

576.7159 570.0373 0.0094 0.8231 0.93

Total area

managed by the

household

2.6264 2.3446 0.1337 0.7086 0.25

Share of total

area used for

cropping

0.8881 0.8521 0.1855 0.5883 0.13

Share of cropped

area dedicated to

cash crops

0.0965 0.1227 -0.2030 0.5781 0.09

Value of

productive assets

101.8320 66.0513 0.1452 3.8705 0.15

63

Table 17: Summary of Balance after matching in Morogoro

Means

Treated

Means

Control

Std. Mean

Diff.

Var.

Ratio

p-value

distance 0.1962 0.1959 0.0036 1.0248

Age of household head 47.4062 47.8877 -0.0287 0.7572 0.71

Education of household

head

5.3895 5.3423 0.0159 0.9639 0.44

Hired labour 0.1314 0.1378 -0.0263 1.2641 0.72

Total area managed by the

household

2.3341 2.0173 0.1841 1.2238 0.12

Share of total area used for

cropping

0.8764 0.8704 0.0286 1.3000 0.97

Share of cropped area

dedicated to cash crops

0.2344 0.2231 0.0417 1.1916 0.81


dedicated to maize

0.5881 0.5811 0.0223 1.3078 0.64

Total labour invested per

land unit

646.5104 707.4235 -

0.0980

0.4708 0.71

Land security index 0.3936 0.4079 -0.0657 1.2461 0.80

Table 18: Summary of Balance before matching in Morogoro

Means

Treated

Means

Control

Std. Mean

Diff.

Var.

Ratio

p-value

distance 0.2149 0.1698 0.4411 2.0307

Age of household head 49.3043 47.7962 0.0900 0.9903 0.50

Education of household

head

5.5062 4.8174 0.2320 0.7796 0.09

Hired labour 0.1295 0.1720 -0.1726 0.8696 0.20

Total area managed by the

household

2.4046 2.0263 0.2197 0.7600 0.11

Share of total area used for

cropping

0.8808 0.8519 0.1385 0.8944 0.30


dedicated to cash crops

0.2174 0.2217 -0.0158 1.1236 0.90


dedicated to maize

0.5932 0.6271 -0.1077 1.2415 0.41

Total labour invested per

land unit

641.1608 656.3226 -0.0244 0.6132 0.86

Land security index 0.4185 0.3715 0.2160 1.9580 0.09

64

Figure 3:Absolute Standardized Mean Differences of matched sample from Dodoma

Figure 4:Distributional Balance for “share of cash crops” covariate before and after matching in

Dodoma

65

Figure 5:Distributional Balance for “percentage of total area used for cropping” covariate after

matching in Dodoma

Figure 6:Absolute Standardized Mean Differences of matched sample from Morogoro

66

Figure 7:Distributional Balance for “education of the household head” covariate before and after

matching in Morogoro

Figure 8: Distributional Balance for “land security indicator” covariate before and after matching

in Morogoro

67

Figure 9:Distributional Balance for “total area managed by the household” covariate after

matching in Morogoro

68


Micro-dosing Adoption in Dodoma, 2013–2016 (Before matching algorithm applied)

Dependent variable: ATT

Mean Difference

of the Matched

Control Group

crop gross margin per ha 0.751 56.834**

(41.543) (22.328)

crop expenditures per ha 17.268 14.844**

(14.618) (5.953)

labour productivity 0.065 -0.044

(0.106) (0.062)

post-harvest loss -1.266 -0.710

(1.250) (0.619)

net household income 40.564 1,357.252***

(181.942) (100.617)

high income fluctuation 0.083 -0.074*

(0.082) (0.038)

has savings 0.113* 0.587***

(0.059) (0.030)

loss of income due to shock 145.927 37.416

(138.566) (44.432)

high severity of shock 0.039 -0.039

(0.071) (0.037)

months to recover after shock 14.187 -10.523***

(19.192) (3.377)

Potential Food Availability index (PFAI) 0.119 -0.004

(0.273) (0.198)

Months of inadequate food provisioning -0.064 -1.000***

(0.514) (0.294)

Food Consumption Score (FCS) -0.222 1.177

(2.437) (1.144)

Coping Strategies Index (CSI) 4.173 -28.677***

(3.006) (1.934)

quantity of applied fertilizer -614.812* 617.224**

(332.265) (299.069)

quantity of animal manure 19.725 -64.000

(123.324) (46.050)

presence of irrigation -0.025 0.016

(0.021) (0.013)

rainfall water use efficiency 0.112 0.591***

(0.075) (0.037)

change in household water consumption -33.833* -18.958**

(19.236) (7.608)

water harvesting 0.560*** -0.397***

(0.069) (0.030)

water use conflict 0.067** -0.128***

(0.033) (0.021)

69

crop performance index -0.004 0.003

(0.015) (0.007)

profitability index -0.060 0.072***

(0.037) (0.020)

stability index -0.105* 0.327***

(0.055) (0.028)

vulnerability to shocks reduction index -0.036 0.063***

(0.025) (0.015)

food security index -0.017 0.114***

(0.023) (0.012)

soil management index -0.008 -0.035***

(0.019) (0.010)

water management index 0.002 0.025***

(0.017) (0.009)

agricultural diversity index -0.035* 0.009

(0.020) (0.010)

economic sustainability index -0.051** 0.116***

(0.020) (0.011)

social sustainability index 0.005 0.002

(0.016) (0.008)

environmental sustainability index -0.014 -0.0003

(0.013) (0.006)

overall sustainability index -0.020 0.039***

(0.012) (0.006)

Note: *p<0.1; **p<0.05; ***p<0.01

70

Table 20:Average Treatment Effect on the Treated (ATT) of Rainwater Harvesting and Fertilizer

Micro-dosing Adoption in Morogoro, 2013–2016 (Before matching algorithm applied)

Dependent variable: ATT

Mean Difference

of the Matched

Control Group

crop gross margin per ha -15.338 -217.204**

(129.892) (84.525)

crop expenditures per ha 89.990 -35.627

(84.194) (76.993)

labour productivity 0.295 -0.559***

(0.249) (0.168)

post-harvest loss 0.828 1.684***

(1.391) (0.540)

net household income -105.797 1,419.753***

(171.051) (96.755)

high income fluctuation -0.033 0.395***

(0.082) (0.036)

has savings 0.025 0.323***

(0.079) (0.035)

loss of income due to shock 301.302 558.606***

(185.991) (53.209)

high severity of shock 0.014 0.464***

(0.080) (0.032)

months to recover after shock 2.820 36.066**

(48.595) (18.271)

Potential Food Availability index (PFAI) -0.177 0.084

(0.345) (0.122)

Months of inadequate food provisioning -1.744** -0.720**

(0.755) (0.324)

Food Consumption Score (FCS) 5.028* -0.231

(2.698) (1.091)

Coping Strategies Index (CSI) -9.685** -7.246***

(4.231) (1.539)

quantity of applied fertilizer -362.841 362.992

(263.536) (263.530)

quantity of animal manure -19.451 -24.389

(46.477) (20.719)

presence of irrigation 0.012 0.031**

(0.036) (0.015)

rainfall water use efficiency -0.100 0.532***

(0.068) (0.046)

change in household water consumption -23.968** 1.720

(11.844) (5.307)

water harvesting 0.404*** -0.013

(0.076) (0.024)

water use conflict 0.026 -0.010

(0.045) (0.016)

71

crop performance index 0.017 -0.094***

(0.025) (0.010)

profitability index 0.044 0.018

(0.059) (0.022)

stability index -0.105* 0.033

(0.062) (0.028)

vulnerability to shocks reduction index 0.002 -0.124***

(0.033) (0.014)

food security index 0.094*** 0.040***

(0.027) (0.013)

soil management index 0.014 0.017**

(0.017) (0.008)

water management index 0.005 0.004

(0.021) (0.008)

agricultural diversity index 0.028 0.067***

(0.019) (0.009)

economic sustainability index -0.011 -0.042***

(0.026) (0.012)

social sustainability index 0.012 -0.020***

(0.015) (0.007)

environmental sustainability index 0.016 0.030***

(0.012) (0.005)

overall sustainability index 0.006 -0.011*

(0.012) (0.006)

Note: *p<0.1; **p<0.05; ***p<0.01

72

Table 21: Average Treatment Effect on the Treated (ATT) of rainwater harvesting and fertilizer

micro-dosing adoption in yield of maize of Morogoro, 2016 - 2013

Dependent variable:

maize_kg_ha_diff

Data before matching Data after matching

ATT 195.168* 199.047*

(101.870) (109.645)

Constant -310.981*** -281.033***

(42.959) (48.806)

Observations 388 323

R2 0.009 0.010

Adjusted R2 0.007 0.007

Residual Std. Error 767.276 (df = 386) 785.466 (df = 321)

F Statistic 3.670* (df = 1; 386) 3.296* (df = 1; 321)

Note: *p<0.1; **p<0.05; ***p<0.01

the impact of rainwater harvesting and fertilizer micro

Documents